Electronic fence-based tracking method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide an electronic fence-based tracking method and device, electronic equipment and storage medium. Based on the obtained target face feature and the preset mapping relationship set, the corresponding target terminal feature is determined. Due to the mapping relationship between the target face feature and the target terminal feature, the spatio-temporal correlation between the two is directly represented. Therefore, compared with the prior art, the correlation between the target face feature and the target terminal feature is higher, thereby improving the tracking accuracy. Meanwhile, the tracking method provided by the embodiments of the present application only needs to match the obtained target face feature with a small number of face features stored in the mapping relationship set, and the target terminal feature can be directly and accurately obtained, thereby effectively improving the correlation efficiency of key data in the tracking process.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular to a tracking method, device, electronic device and storage medium based on electronic fences. Background Technology

[0002] The International Mobile Subscriber Identity (IMSI) is a unique identifier used in cellular networks to distinguish different terminals. Due to its uniqueness, the IMSI of a target terminal can be used for targeted eavesdropping and tracking. Therefore, in the field of criminal investigation, the IMSI of the target terminal carried by the suspect is often used to eavesdrop on and track the target terminal, thereby analyzing the suspect's behavior and uncovering their whereabouts.

[0003] Against this backdrop, in order to enhance social security, public security big data systems typically deploy corresponding "electronic fences" based on IMSI (Internet Security Identity). These fences provide real-time warnings when a suspect carrying a target terminal comes into contact with the network, and continuously track the target terminal based on its IMSI, thereby improving the efficiency of case detection.

[0004] Under the relevant technologies, the deployment of "electronic fences" first requires binding and associating the IMSI of the target terminal carried by the suspect with the suspect's own facial data. However, the specific implementation plan still has the following drawbacks:

[0005] 1. Low correlation efficiency.

[0006] Due to the widespread availability of smart devices, the sheer number of them in urban areas results in an enormous volume of IMSI data. Under current technology, the only way to identify a target device's IMSI is to painstakingly sift through this massive amount of data using a trained model. This process is time-consuming and inefficient.

[0007] 2. Low correlation accuracy.

[0008] Under related technologies, multiple training models are used to determine the correlation between the facial data of the target suspect and the IMSI of the target terminal through deep fitting. However, the correlation obtained in this way is not reliable and has low accuracy. Summary of the Invention

[0009] This application provides a tracking method, device, electronic device, and storage medium based on electronic fences to solve the problems of low efficiency and low accuracy in associating key data in related technologies.

[0010] In a first aspect, embodiments of this application provide a tracking method based on an electronic fence, comprising:

[0011] Obtain the first candidate feature data set of each candidate object collected; wherein each first candidate feature data set includes at least: the facial features of the corresponding candidate object, the facial feature acquisition time, and the facial feature acquisition location.

[0012] Obtain the second candidate feature data set of each candidate terminal collected; wherein each second candidate feature data set includes at least: the terminal feature of the corresponding candidate terminal, the terminal feature collection time, and the terminal feature collection location.

[0013] Based on the acquisition time and location of each facial feature in the first candidate feature data set, and the acquisition time and location of each terminal feature in the second candidate feature data set, the spatiotemporal correlation between each acquired facial feature and the corresponding terminal feature is determined.

[0014] The spatiotemporal correlation between each facial feature and its corresponding terminal feature is input into a preset correlation model to obtain the mapping relationship between each facial feature and its corresponding terminal feature.

[0015] Secondly, embodiments of this application also provide a tracking device based on an electronic fence, comprising:

[0016] The acquisition module is used to acquire the target facial image of the target object through the camera device, obtain the target object's current target location information, and perform image processing on the target facial image to obtain the target facial features of the target object.

[0017] The determination module is used to determine the target terminal features corresponding to the target facial features based on a preset mapping relationship set. The mapping relationship set includes: the association between the facial features of each candidate object and the terminal features of each candidate terminal, and each association is determined based on the facial feature acquisition location and facial feature acquisition time of the candidate object, as well as the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal.

[0018] The tracking module is used to track the target object based on the obtained target terminal characteristics when the target location information determines that the target object's movement trajectory exceeds the preset electronic fence range.

[0019] In an optional embodiment, before acquiring the target facial features of the target object through the camera device, the acquisition module is further configured to:

[0020] Obtain the first candidate feature data set of each candidate object collected; wherein each first candidate feature data set includes at least: the facial features of the corresponding candidate object, the facial feature acquisition time, and the facial feature acquisition location.

[0021] Obtain the second candidate feature data set of each candidate terminal collected; wherein each second candidate feature data set includes at least: the terminal feature of the corresponding candidate terminal, the terminal feature collection time, and the terminal feature collection location.

[0022] Based on the acquisition time and location of each facial feature in the first candidate feature data set, and the acquisition time and location of each terminal feature in the second candidate feature data set, the spatiotemporal correlation between each acquired facial feature and the corresponding terminal feature is determined.

[0023] The spatiotemporal correlation between each facial feature and its corresponding terminal feature is input into a preset correlation model to obtain the mapping relationship between each facial feature and its corresponding terminal feature.

[0024] In an optional embodiment, when acquiring the first set of candidate feature data for each candidate object, the acquisition module is specifically used for:

[0025] Using a time window overlay algorithm, based on a preset time range, the first candidate feature data set of each candidate object is selected from the first candidate feature data obtained by feature collection for each object.

[0026] In an optional embodiment, when acquiring the second candidate feature data set of each candidate terminal, the acquisition module is specifically used for:

[0027] Using a time window overlay algorithm, based on a preset time range, the second candidate feature data set of each candidate terminal is selected from the second candidate feature data obtained by feature collection for each terminal.

[0028] In an optional embodiment, when determining the spatiotemporal correlation between each collected facial feature and its corresponding terminal feature based on the facial feature acquisition time and location in the obtained first candidate feature data set, and the terminal feature acquisition time and location in the obtained second candidate feature data set, the acquisition module is specifically used for:

[0029] Based on the acquisition time of each facial feature in the first candidate feature data set and the acquisition time of each terminal feature in the second candidate feature data set, the temporal correlation between each facial feature and its corresponding terminal feature is determined.

[0030] Based on the facial feature acquisition locations in the first candidate feature data set and the terminal feature acquisition locations in the second candidate feature data set, the spatial association between each facial feature and its corresponding terminal feature is determined.

[0031] Based on the obtained temporal and spatial correlations, the spatiotemporal correlations between each collected facial feature and the corresponding terminal feature are determined.

[0032] In an optional embodiment, when determining the temporal correlation between each facial feature and its corresponding terminal feature based on the acquisition time of each facial feature in the obtained first candidate feature data set and the acquisition time of each terminal feature in the obtained second candidate feature data set, the acquisition module is specifically used for:

[0033] For each facial feature acquisition time and the corresponding terminal feature acquisition time, perform the following operations respectively:

[0034] Determine the time difference between a facial feature acquisition time and the corresponding terminal feature acquisition time.

[0035] If the obtained time difference is less than the preset time difference threshold, the temporal correlation between facial features and corresponding terminal features is determined based on the time difference.

[0036] If the obtained time difference is not less than the preset time difference threshold, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the second candidate feature data set.

[0037] In an optional embodiment, when determining the spatial association between each facial feature and its corresponding terminal feature based on the facial feature acquisition locations in the obtained first candidate feature data set and the terminal feature acquisition locations in the obtained second candidate feature data set, the acquisition module is specifically used for:

[0038] For each facial feature acquisition location and the corresponding terminal feature acquisition location, perform the following operations respectively:

[0039] Obtain the physical distance between a facial feature acquisition location and the corresponding terminal feature acquisition location.

[0040] If the obtained physical distance is less than the preset distance threshold, the spatial relationship between facial features and corresponding terminal features is determined based on the physical distance.

[0041] If the obtained physical distance is not less than the preset standard distance, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the obtained second candidate feature data set.

[0042] In an optional embodiment, when inputting the spatiotemporal correlation between the obtained facial features and corresponding terminal features into a preset correlation model to obtain the mapping relationship between each facial feature and its corresponding terminal feature, the acquisition module is specifically used for:

[0043] For each facial feature and its corresponding terminal feature, perform the following operations respectively:

[0044] Input the spatiotemporal correlation between a facial feature and its corresponding terminal feature into a preset correlation model to obtain the confidence weights corresponding to the spatiotemporal correlation.

[0045] If the obtained confidence weight is less than the preset confidence weight threshold, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the obtained second candidate feature data set.

[0046] If the obtained confidence weight is not less than the preset confidence weight threshold, then the mapping relationship between facial features and corresponding terminal features is determined based on the obtained spatiotemporal correlation.

[0047] Thirdly, embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, the processor enables the processor to implement any of the electronic fence-based tracking methods described in the first aspect above.

[0048] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the electronic fence-based tracking method of the first aspect.

[0049] The target facial image of the target object is acquired by a camera device, and the current target location information of the target object is obtained. The target facial image is then processed to obtain the target facial features of the target object.

[0050] Based on a preset set of mapping relationships, the target terminal features corresponding to the target facial features are determined. The set of mapping relationships includes the association between the facial features of each candidate object and the terminal features of each candidate terminal. Each association is determined based on the facial feature acquisition location and facial feature acquisition time of the candidate object, as well as the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal.

[0051] When the target object's movement trajectory is determined to be outside the preset electronic fence range based on the target location information, the target object is tracked based on the target terminal characteristics.

[0052] In this embodiment, image processing is performed on the acquired target facial image to obtain the target facial features of the target object. Based on the target facial features and a preset mapping relationship set, the corresponding target terminal features are determined. Since the mapping relationship set contains the correlation between the facial features of each candidate object and the terminal features of each candidate terminal, and each mapping relationship is determined only by the facial feature acquisition location and facial feature acquisition time of the candidate object, as well as the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal, the correlation between each mapping relationship in the mapping relationship set is higher. That is, the correlation between the obtained target facial features and target terminal features is higher, thereby further improving the tracking accuracy.

[0053] Furthermore, as can be seen from the above process, this method only needs to match the obtained target facial features with a small number of facial features stored in the mapping relationship set to directly and accurately obtain the target terminal features, thereby effectively improving the correlation efficiency of key data during the tracking process, making the obtained action trajectory more complete, and providing more time for action to discover the target object's location. Attached Figure Description

[0054] Figure 1 An architecture diagram of an electronic fence system provided in an embodiment of this application;

[0055] Figure 2 Example diagram of facial feature acquisition provided in the embodiments of this application;

[0056] Figure 3 This is a schematic diagram of terminal feature acquisition provided in an embodiment of this application;

[0057] Figure 4 A flowchart illustrating the process of setting up relevant mapping relationships for the management platform provided in this application embodiment;

[0058] Figure 5 Example diagrams of facial feature acquisition provided in this application embodiment;

[0059] Figure 6 Example diagrams illustrating the features of the data acquisition terminal provided in the embodiments of this application;

[0060] Figure 7 A schematic diagram illustrating the process of obtaining spatiotemporal correlations provided in an embodiment of this application;

[0061] Figure 8 A schematic diagram illustrating the process of obtaining time-related relationships provided in an embodiment of this application;

[0062] Figure 9 A schematic diagram illustrating the process of obtaining spatial relationships provided in an embodiment of this application;

[0063] Figure 10 A schematic diagram illustrating the process of obtaining spatiotemporal correlations provided in an embodiment of this application;

[0064] Figure 11 A flowchart illustrating the tracking method based on an electronic fence provided in this application embodiment;

[0065] Figure 12 Example diagrams of facial features of the target object provided in this application embodiment;

[0066] Figure 13 Example diagrams for tracking the movement trajectory of target objects provided in embodiments of this application;

[0067] Figure 14 A schematic diagram of a tracking device based on an electronic fence provided for an embodiment of this application;

[0068] Figure 15 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0069] To address the issue of low efficiency in associating key data in existing technologies, this application embodiment filters the collected key data based on time window overlay, thereby ensuring the availability of key data. Furthermore, to address the issue of low accuracy in associating key data in related technologies, this application embodiment obtains corresponding spatiotemporal relationships by analyzing the collection time and location information of the key data. Since the spatiotemporal relationships are determined solely by the collection time and location information and are not affected by other external factors, this spatiotemporal relationship is used as a mapping relationship between key data. Compared to the association relationships obtained through deep fitting in related technologies, the mapping relationship determined in this way has higher accuracy, thus avoiding the problem of subsequent erroneous transmission due to insufficient accuracy in the association relationship.

[0070] The preferred embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0071] See Figure 1 As shown in the embodiments of this application, within the preset electronic fence area, various camera devices (such as cameras, webcams, etc.) and various base stations supporting electronic fences (hereinafter referred to as electronic fence base stations) are deployed, and each camera device and each electronic fence base station has a data transmission channel with the management platform.

[0072] The management platform, through various camera devices, can acquire facial features of each object, the time of facial feature acquisition, and the location of facial feature acquisition. For example:

[0073] See Figure 2 As shown, taking road A within the electronic fence as an example, a camera A with facial feature acquisition function is deployed on road A. The management platform can acquire the facial features of object A who passes through point A-1 on road A at 12:00 by using camera A deployed at point A-1 on road A, and record the time 12:00 as the facial feature acquisition time, and record the location corresponding to point A-1 as the facial feature acquisition location.

[0074] The management platform can obtain the terminal characteristics, terminal characteristic collection time, and terminal characteristic collection location of each terminal through each base station. For example:

[0075] See Figure 3 As shown, taking road A as an example, if a base station B is deployed on road A, the management platform can obtain the terminal features of terminal B located in the range B with B-1 as the center and a preset detection distance as the radius at 12:05 through the base station B deployed at point B-1 on road A (assuming that terminal B is carried by object B at this time). The platform records the time 12:05 as the terminal feature acquisition time and the location corresponding to point B-1 as the terminal feature acquisition location.

[0076] See Figure 4 As shown, in the preprocessing stage, in this embodiment of the application, the management platform obtains the mapping relationship set in the following specific way:

[0077] Step 41: Obtain the first candidate feature data set of each candidate object collected; wherein each first candidate feature data set includes at least: the facial features of the corresponding candidate object, the facial feature acquisition time, and the facial feature acquisition location.

[0078] In this embodiment of the application, the management platform obtains the first candidate feature data collected by each camera device within a preset time range through the data transmission channel between the management platform and each camera device, including:

[0079] Facial features of the corresponding candidate;

[0080] The recorded acquisition time for each facial feature;

[0081] The locations where facial features are captured are determined based on the deployment location of the camera equipment.

[0082] For example: See Figure 5 As shown, a fixed camera A is deployed at point A-1 on road A, with a half-angle of 180° within a set range. The management platform obtains the facial features of each candidate object that passes through the 180° range of point A-1, as well as the recorded facial feature acquisition time, through the data transmission channel. The management platform also obtains the facial feature acquisition position of each candidate object based on the deployment position of camera A (i.e., point A-1).

[0083] Referring to Table 1 below, assuming that camera A collects the following set of first candidate feature data within a preset time range (24 hours):

[0084] Table 1

[0085] Candidates Facial features Facial feature acquisition time Facial feature acquisition location Candidate 1 F(1) 10:00 (104.78734，36.93642) Candidate 2 F(2) 11:30 (104.78735，36.93642) Candidate 3 F(3) 12:30 (104.78744，36.93642) Candidate 4 F(4) 13:00 (104.78735，36.93652) Candidate 5 F(5) 14:30 (104.78737，36.93643)

[0086] In one optional embodiment, to ensure the precision of data collection, a time window overlay algorithm is used to select the first set of candidate feature data for each candidate object from the candidate feature data obtained by feature collection for each object within a preset time range.

[0087] Specifically, within a preset time range, the camera device may collect the first candidate feature data of the same candidate object multiple times. If the management platform directly obtains the mapping set based on the above data, it is easy to generate data redundancy and affect the association efficiency. To solve the above problem, the management platform adopts a time window superposition algorithm to filter the first candidate feature data collected by the camera device based on the number of times each first candidate feature data is collected within a unit of time.

[0088] For example, referring to Table 2, assuming that within a preset time range (24 hours), camera A performs feature acquisition on each object, the obtained first candidate feature data are as follows:

[0089] Table 2

[0090] Facial feature acquisition time object Facial features Facial feature acquisition location 0:00 object a F(a) (104.78733，36.93642) … … … … 9:00 object b F(b) (104.78733，36.93642) 9:10 object b F(b) (104.78732，36.93642) 9:20 object b F(b) (104.78731，36.93642) 10:00 object c F(c) (104.78734，36.93642) 11:30 object d F(c) (104.78735，36.93642) 11:40 object d F(c) (104.78735，36.93641) 12:00 object e F(e) (104.78744，36.93642) … … … … 15:00 object d F(d) (104.78731，36.93642) 15:10 object d F(d) (104.78731，36.93642) … … … … 24:00 object f F(f) (104.78735，36.93652)

[0091] The management platform obtains the first candidate feature data as shown in Table 2. Among them, at 9:00, 9:10 and 9:20, the management platform obtains the first candidate feature data of the same object (object b) multiple times. Since the above three time points are all within the same unit of time, the management platform uses the time window superposition algorithm to remove the first candidate feature data of object b.

[0092] Meanwhile, as shown in Table 2, at 11:00, 11:40, 15:00, and 15:10, the management platform acquired the first candidate feature data of object d multiple times. Specifically, within two time units (11:00-12:00 and 15:00-16:00), the first candidate feature data of the same object (object d) was acquired twice. Therefore, the management platform adopted the time window overlay algorithm to remove the first candidate feature data of object d.

[0093] In real-world scenarios, camera equipment is typically deployed in fixed areas with high object flow density, such as street intersections. Therefore, if the object being collected is located near the street intersection for an extended period, the first candidate feature data of the relevant object is likely to be collected multiple times. Using the method described above, the first candidate feature data set of each candidate object can be selected from the first candidate feature data obtained by collecting features from each object, thereby avoiding unnecessary association of non-critical data in subsequent processes and preventing additional computational load.

[0094] In an optional embodiment, to simplify the process of obtaining facial feature acquisition locations and improve data acquisition efficiency, the deployment location of the camera device is used as the facial feature acquisition location for each candidate object. For example, assuming the deployment location of camera A is (104.78734, 36.93642), then the deployment location (104.78734, 36.93642) is used as the facial feature acquisition location for each candidate object. The first candidate feature data set collected by camera A within a preset time range (24 hours) is shown in Table 3 below.

[0095] Table 3

[0096] Candidates Facial features Facial feature acquisition time Facial feature acquisition location Candidate 1 F(1) 10:00 (104.78734，36.93642) Candidate 2 F(2) 11:30 (104.78734，36.93642) Candidate 3 F(3) 12:30 (104.78734，36.93642) Candidate 4 F(4) 13:00 (104.78734，36.93642) Candidate 5 F(5) 14:30 (104.78734，36.93642)

[0097] Step 42: Obtain the second candidate feature data set of each candidate terminal; wherein each second candidate feature data set includes at least: the terminal feature of the corresponding candidate terminal, the terminal feature acquisition time, and the terminal feature acquisition location.

[0098] In this embodiment of the application, the management platform, through the data transmission channel with each power grid base station, uses network gateway mapping and other technical means to obtain the second candidate feature data collected by each power grid base station within a preset time range, including:

[0099] Terminal characteristics of the corresponding candidate terminals, such as the IMSI of the corresponding candidate terminals;

[0100] The time of feature acquisition for each recorded terminal;

[0101] Based on the deployment location of the base station, the location of each terminal feature was obtained.

[0102] For example: See Figure 6 As shown, a fixed mobile base station B is deployed at point B-1 on road A. Its collection half-angle is 360° within a set range. The management platform obtains the IMSI of each candidate terminal and the facial feature collection time recorded by each of them through the 360° range of point B-1, collected by the mobile base station B, through data transmission. The management platform also obtains the terminal feature collection position corresponding to each candidate terminal based on the deployment position of the mobile base station B (i.e., point B-1).

[0103] Referring to Table 4 below, assuming that the second candidate feature data set collected by base station B within a preset time range (24 hours) is as follows:

[0104] Table 4

[0105] Candidate Terminal Terminal characteristics Terminal feature acquisition time Terminal feature acquisition location Candidate Terminal 1 IMSI(1) 10:05 (104.78754，36.93642) Candidate Terminal 2 IMSI(2) 11:31 (104.78755，36.93642) Candidate Terminal 3 IMSI(3) 13:01 (104.78754，36.93642) Candidate Terminal 4 IMSI(4) 14:05 (104.78755，36.93652) Candidate Terminal 5 IMSI(5) 14:35 (104.78757，36.93643)

[0106] In one optional embodiment, to ensure the precision of data collection, the management platform adopts a time window overlay algorithm. Based on a preset time range, it selects the second candidate feature data set of each candidate terminal from the second candidate feature data obtained by feature collection for each terminal.

[0107] Specifically, within a preset time range, the base station may acquire the second candidate feature data of the same candidate terminal multiple times. If the management platform directly obtains the mapping set based on the above data, it is easy to generate data redundancy and affect the association efficiency. To solve the above problem, the management platform adopts a time window superposition algorithm to filter the acquired second candidate feature data based on the number of times each second candidate feature data is collected within a unit of time.

[0108] For example, referring to Table 5, assuming that within a preset time range (24 hours), base station B collects features from each terminal, the obtained second candidate feature data are as follows:

[0109] Table 5

[0110]

[0111]

[0112] The management platform obtains the second candidate feature data as shown in Table 5. At 9:00, 9:10, and 9:20, the management platform obtains the second candidate feature data of the same terminal (terminal b) multiple times. Since the above three time points are all within the same unit of time, the management platform uses the time window superposition algorithm to remove the second candidate feature data of object b.

[0113] Meanwhile, as shown in Table 2, at 11:00, 11:40, 15:00, and 15:10, the management platform acquired the first candidate feature data of terminal d multiple times. Specifically, the management platform acquired the second candidate feature data of the same terminal (terminal d) twice in two time units (11:00-12:00 and 15:00-16:00). Therefore, the management platform used the time window overlay algorithm to remove the second candidate feature data of terminal d.

[0114] In real-world scenarios, base stations are often deployed in fixed areas with high terminal mobility. Therefore, for each second candidate feature data obtained by the base station from feature collection of each terminal, a time window overlay algorithm is also needed to filter out the second candidate feature data set of each candidate terminal. This avoids unnecessary association of non-critical data in subsequent processes, which would cause additional computational load.

[0115] In an optional embodiment, to simplify the process of obtaining terminal feature collection locations and improve data acquisition efficiency, the deployment location of the base station is used as the terminal feature collection location for each candidate terminal. For example, assuming the deployment location of base station B is (104.78754, 36.93642), then the deployment location (104.78754, 36.93642) is used as the terminal feature collection location for each candidate terminal. The set of second candidate feature data collected by base station B within a preset time range (24 hours) is shown in Table 6 below.

[0116] Table 6

[0117] Candidate Terminal Terminal characteristics Terminal feature acquisition time Terminal feature acquisition location Candidate Terminal 1 IMSI(1) 10:05 (104.78754，36.93642) Candidate Terminal 2 IMSI(2) 11:31 (104.78754，36.93642) Candidate Terminal 3 IMSI(3) 13:01 (104.78754，36.93642) Candidate Terminal 4 IMSI(4) 14:05 (104.78754，36.93642) Candidate Terminal 5 IMSI(5) 14:35 (104.78754，36.93642)

[0118] Based on the above-mentioned time window overlay method, the collected key data is prioritized for filtering, thereby ensuring the availability and accuracy of the key data. This method also has good adaptability. For example, in this embodiment, the time window is set to hours / days, while in some special scenarios, the time window can be set to days / months to more accurately filter out key data.

[0119] Step 43: Based on the facial feature acquisition time and location in the first candidate feature data set, and the terminal feature acquisition time and location in the second candidate feature data set, determine the spatiotemporal correlation between each acquired facial feature and the corresponding terminal feature.

[0120] Since there is an inevitable temporal and spatial correlation between facial features and corresponding terminal features, the spatiotemporal relationship between each facial feature and its corresponding terminal feature can be determined by recording relevant data.

[0121] In one alternative embodiment, see [reference] Figure 7 As shown, the spatiotemporal correlation between each collected facial feature and its corresponding terminal feature is determined through the following methods:

[0122] Step 431: Based on the acquisition time of each facial feature in the first candidate feature data set and the acquisition time of each terminal feature in the second candidate feature data set, determine the temporal correlation between each facial feature and the corresponding terminal feature.

[0123] Specifically, in real-world scenarios, there is an inherent temporal correlation between the facial feature acquisition time of an object and the terminal feature acquisition time of the terminal carried by the same object, unaffected by any external factors. Therefore, the management platform determines the temporal correlation between facial features and corresponding terminal features based on the acquired facial feature acquisition time and the corresponding terminal feature acquisition time.

[0124] Taking the first candidate feature data set shown in Table 1 and the second candidate feature data set shown in Table 4 as examples, the determined time-related relationships are shown in Table 7 below:

[0125] Table 7

[0126] Facial features Terminal characteristics F(1) IMSI(1) F(2) IMSI(2) F(4) IMSI(3) F(5) IMSI(5)

[0127] In one alternative embodiment, see [reference] Figure 8 As shown, the temporal correlation between each collected facial feature and its corresponding terminal feature is determined through the following methods:

[0128] Step 4311: For each facial feature acquisition time and the corresponding terminal feature acquisition time, perform the following operations respectively: Determine the time difference between a facial feature acquisition time and the corresponding terminal feature acquisition time;

[0129] Taking the facial feature acquisition time of 10:00 corresponding to facial feature F(1) in Table 1 as an example, the corresponding terminal feature acquisition time is the terminal feature acquisition time of 10:05 corresponding to terminal feature IMSI(1) in Table 4. Then the time difference between facial feature F(1) and the corresponding terminal feature IMSI(1) is 5 minutes.

[0130] Similarly, taking the facial feature acquisition time of 12:30 corresponding to facial feature F(3) in Table 1 as an example, the corresponding terminal feature acquisition time is the terminal feature acquisition time of 13:01 corresponding to terminal feature IMSI(3) in Table 4. Then the time difference between facial feature F(3) and the corresponding terminal feature IMSI(3) is 31 minutes.

[0131] Using the above method, the management platform can obtain the time difference between the acquisition time of each facial feature in Table 1 and the corresponding terminal feature acquisition time in Table 4, which will not be elaborated further here.

[0132] Step 4312: Determine whether the obtained time difference is less than the preset time difference threshold. If yes, proceed to step 4313; otherwise, proceed to step 4314.

[0133] In real-world scenarios, since camera equipment and mobile base stations are usually deployed in different locations, facial feature acquisition time and terminal feature acquisition time often have a sequential relationship. Taking any object whose facial features are acquired as an example (hereinafter referred to as object x), considering the possible time difference between the facial feature acquisition time of object x and the terminal feature acquisition time of the terminal carried by the same object, the management platform can optionally determine the corresponding time correlation relationship based on whether the obtained time difference is less than a preset time difference threshold.

[0134] Step 4313: Based on the obtained time difference, determine the temporal correlation between a facial feature and the corresponding terminal feature.

[0135] Specifically, if the management platform determines that the obtained time difference is less than the preset time difference threshold, it means that within the time range corresponding to the time difference, the facial features of object x and the terminal features of the terminal x carried by object x are obtained simultaneously. Based on the obtained time difference, the management platform determines the temporal correlation between the obtained facial features and the corresponding terminal features.

[0136] For example, assuming the preset time difference threshold is 10 minutes, and the time difference between the obtained facial feature F(1) and the corresponding terminal feature IMSI(1) is 5 minutes, the management platform determines that the obtained time difference is less than the preset time difference threshold. Based on the obtained time difference, the management platform determines the time correlation between the obtained facial feature F(1) and the corresponding terminal feature IMSI(1).

[0137] Step 4314: Remove the corresponding facial features from the obtained first candidate feature data set, and remove the corresponding terminal features from the obtained second candidate feature data set.

[0138] Specifically, if the management platform determines that the obtained time difference is not less than the preset time difference threshold, it means that within the time range corresponding to the time difference, the possibility of simultaneously obtaining the facial features of object x and the terminal features of the terminal x carried by object x is small.

[0139] For example, assuming the preset time difference threshold is 10 minutes, and the time difference between the obtained facial feature F(3) and the corresponding terminal feature IMSI(3) is 31 minutes, the management platform determines that the obtained time difference is not less than the preset time difference threshold, indicating that the possibility of obtaining facial feature F(3) and the corresponding terminal feature is small. Therefore, the management platform removes the corresponding facial feature F(3) from the first candidate feature data set shown in Table 1 and removes the corresponding terminal feature IMSI(3) from the second candidate feature data set shown in Table 3.

[0140] Step 432: Based on the facial feature acquisition locations in the first candidate feature data set and the terminal feature acquisition locations in the second candidate feature data set, determine the spatial association between each acquired facial feature and its corresponding terminal feature.

[0141] Specifically, in real-world scenarios, there is an inherent spatial correlation between the facial feature acquisition location of an object and the terminal feature acquisition location of the terminal carried by the same object, without being affected by any external factors. Therefore, the management platform determines the spatial correlation between facial features and corresponding terminal features based on the acquired facial feature acquisition location and the corresponding terminal feature acquisition location.

[0142] Taking the first candidate feature data set shown in Table 1 and the second candidate feature data set shown in Table 4 as examples, the determined spatial relationships are shown in Table 8 below:

[0143] Table 8

[0144] Facial features Terminal characteristics F(1) IMSI(1), ..., IMSI(5) F(2) IMSI(1), ..., IMSI(5) F(3) IMSI(1), ..., IMSI(5) F(4) IMSI(1), ..., IMSI(5) F(5) IMSI(1), ..., IMSI(5)

[0145] In one alternative embodiment, see [reference] Figure 9 As shown, the spatial association between each collected facial feature and its corresponding terminal feature is determined through the following methods:

[0146] Step 4321: For each obtained facial feature acquisition location and corresponding terminal feature acquisition location, perform the following operations respectively: obtain the physical distance between a facial feature acquisition location and the corresponding terminal feature acquisition location.

[0147] Optionally, the physical distance between a facial feature acquisition location and the corresponding terminal feature acquisition location can be obtained by using the latitude and longitude distance difference method.

[0148] For example, in Table 1, the facial feature acquisition location of facial feature F(1) is (104.78734, 36.93642), and in Table 3, the terminal feature acquisition location of terminal feature IMSI(1) is (104.78754, 36.93642). Using the latitude and longitude distance difference method, the physical distance between F(1) and IMSI(1) is 20m.

[0149] In an optional embodiment, if in the above steps, the deployment location of the camera device is used as the facial feature acquisition location of each candidate object, and the deployment location of the power grid base station is used as the terminal feature acquisition location of each candidate terminal, then in step 4321, the physical distance between each facial feature acquisition location and the corresponding terminal feature acquisition location is obtained based on the deployment location of the camera device and the deployment location of the power grid base station.

[0150] For example, assuming the camera equipment is deployed at (104.78734, 36.93642) and the base station is deployed at (104.78754, 36.93642), the physical distance between the two is obtained as 20m by using the latitude and longitude distance difference method. Based on the obtained physical distance of 20m between the camera equipment and the base station, the management equipment sets the physical distance between each facial feature acquisition location and the corresponding terminal feature acquisition location to be 20m.

[0151] Step 4322: Determine whether the obtained physical distance is less than the preset distance threshold. If yes, proceed to step 4323; otherwise, proceed to step 4324.

[0152] In real-world scenarios, since camera equipment and base stations are usually deployed in different locations, and the objects and terminals may be mobile, there is often a physical distance between the facial feature acquisition location and the terminal feature acquisition location. Taking any object whose facial features are acquired as an example (hereinafter referred to as object x), considering the possible physical distance between the facial feature acquisition location of object x and the terminal feature acquisition location of the terminal carried by the same object, the management platform may optionally determine the corresponding spatial association relationship based on whether the obtained physical distance is less than a preset distance threshold.

[0153] Step 4323: Based on the obtained physical distance, determine the spatial relationship between facial features and corresponding terminal features.

[0154] Specifically, if the management platform determines that the obtained physical distance is less than the preset distance threshold, it means that the corresponding camera device and the corresponding base station are close or adjacent in terms of deployment location. Based on the obtained physical distance, the management platform determines the spatial correlation between the obtained facial features and the corresponding terminal features.

[0155] For example, assuming the preset distance threshold is 50m, and the physical distance between the facial feature acquisition location (104.78734, 36.93642) corresponding to the obtained facial feature F(1) and the terminal feature acquisition location (104.78754, 36.93642) corresponding to the terminal feature IMSI(1) is 20m, then the management platform determines that the obtained physical distance is less than the preset distance threshold, and based on the obtained physical distance, determines the spatial association relationship between the obtained facial feature F(1) and the corresponding terminal feature IMSI(1).

[0156] Step 4324: Remove the corresponding facial features from the obtained first candidate feature data set, and remove the corresponding terminal features from the obtained second candidate feature data set.

[0157] Specifically, if the management platform determines that the obtained physical distance is not less than the preset distance threshold, it means that the corresponding camera device and the corresponding base station are far apart, and then the two may not have simultaneously obtained the facial features of object x and the terminal features of the terminal x carried by object x.

[0158] For example, assuming the preset distance threshold is 10m, and the physical distance between the obtained facial feature F(1) and the corresponding terminal feature IMSI(1) is 20m, the management platform determines that the obtained physical distance is not less than the preset distance threshold, and removes the facial feature F(1) from the first candidate feature data set shown in Table 1, and removes the corresponding terminal feature IMSI(1) from the corresponding second candidate feature data set shown in Table 4.

[0159] Step 433: Based on the obtained temporal and spatial correlations, determine the spatiotemporal correlation between each collected facial feature and the corresponding terminal feature.

[0160] Optionally, the spatiotemporal relationship between facial features and corresponding terminal features can be determined based on whether there is both a temporal and spatial correlation between the facial features and the corresponding terminal features.

[0161] Taking the various temporal relationships shown in Table 7 and the various spatial relationships shown in Table 8 as examples, the facial feature F(1) and the terminal feature IMSI(1) have both temporal and spatial relationships, so the management platform determines that the facial feature F(1) and the terminal feature IMSI(1) have a spatiotemporal relationship. However, the facial feature F(3) and the IMSI(3) have a spatial relationship but no temporal relationship, so the management platform determines that the facial feature F(3) and the IMSI(3) do not have a spatiotemporal relationship.

[0162] Based on the above method, the spatiotemporal correlation between each collected facial feature and the corresponding terminal feature can be obtained. Taking the temporal correlation shown in Table 7 and the spatial correlation shown in Table 8 as examples, the obtained spatiotemporal correlation is shown in Table 9 below:

[0163] Table 9

[0164] Facial features Terminal characteristics F(1) IMSI(1) F(2) IMSI(2) F(4) IMSI(3) F(5) IMSI(5)

[0165] Step 44: Input the spatiotemporal correlation between each facial feature and the corresponding terminal feature into the preset correlation model to obtain the mapping relationship between each facial feature and the corresponding terminal feature.

[0166] In real-world scenarios, to avoid the randomness of data collection, an association model is set up based on the historical collection frequency and time of each facial feature, and the confidence score of each spatiotemporal association is calculated.

[0167] Referring to Table 10 below, the management platform inputs the various spatiotemporal relationships shown in Table 9 into the preset association model, thus determining the mapping relationship between each facial feature and its corresponding terminal feature as follows:

[0168] Table 10

[0169] Facial features Terminal characteristics F(1) IMSI(1) F(2) IMSI(2) F(4) IMSI(3) F(5) IMSI(5)

[0170] In one alternative embodiment, see [reference] Figure 10 As shown, the mapping relationship between each collected facial feature and its corresponding terminal feature is determined in the following ways:

[0171] Step 441: For each facial feature and its corresponding terminal feature, perform the following operations respectively: input the spatiotemporal correlation between a facial feature and its corresponding terminal feature into a preset correlation model to obtain the confidence weight corresponding to the spatiotemporal correlation.

[0172] Specifically, to further ensure the accuracy of the mapping relationship, an association model is applied, and historical data is statistically analyzed to calculate the confidence score between various spatiotemporal associations. Optional factors for calculating the confidence score include: facial features, terminal features, facial feature acquisition time, terminal feature acquisition time, facial feature acquisition location, and terminal feature acquisition location.

[0173] For example, by training an association model using historical data collected from camera equipment and cell tower base stations, the management platform inputs the obtained spatiotemporal relationships into the trained association model. The association model then calculates the confidence weights of each facial feature and its corresponding terminal feature in sequence, as shown in Table 11 below:

[0174] Table 11

[0175] Spatiotemporal correlation Confidence weights F(1)-IMSI(1) 2 F(2)-IMSI(2) 1 F(4)-IMSI(3) 3 F(5)-IMSI(5) 4

[0176] Step 442: Determine whether the obtained confidence weight is less than the preset confidence weight threshold. If yes, proceed to step 443; otherwise, proceed to step 444.

[0177] In this embodiment of the application, the obtained confidence weight is used to represent the reliability of the spatiotemporal correlation. Therefore, the management platform needs to filter each spatiotemporal correlation based on the obtained confidence weight to further ensure the accuracy of the mapping relationship.

[0178] Step 443: Remove the corresponding facial features from the obtained first candidate feature data set, and remove the corresponding terminal features from the obtained second candidate feature data set.

[0179] Specifically, if the management platform determines that the obtained confidence weight is less than the preset confidence weight threshold, it means that the spatiotemporal correlation between the obtained facial features and the corresponding terminal features is not sufficient to accurately represent the mapping relationship between the facial features and the corresponding terminal features. In this case, the management platform removes the corresponding facial features from the obtained first candidate feature data set and removes the corresponding terminal features from the obtained second candidate feature data set.

[0180] Taking the various spatiotemporal correlations obtained in Table 11 above as an example, assuming that the preset confidence weight threshold is 2, the management device determines that the spatiotemporal correlation F(2)-IMSI(2) is less than the preset confidence weight threshold, and removes the corresponding facial feature F(2) from the obtained first candidate feature data set, and removes the corresponding terminal feature IMSI(2) from the obtained second candidate feature data set.

[0181] Step 444: Based on the obtained spatiotemporal correlation, determine the mapping relationship between facial features and corresponding terminal features.

[0182] Specifically, if the management platform determines that the obtained confidence weight is not less than the preset confidence weight threshold, it means that the spatiotemporal correlation between the obtained facial features and the corresponding terminal features is sufficient to accurately represent the mapping relationship between the facial features and the corresponding terminal features. Based on the obtained spatiotemporal correlation, the management platform determines the mapping relationship between the facial features and the corresponding terminal features.

[0183] Taking the various spatiotemporal correlations obtained in Table 11 as examples, assuming that the preset confidence weight threshold is 2, the management device determines that the spatiotemporal correlations F(1)-IMSI(1), F(4)-IMSI(3), and F(5)-IMSI(5) are all not less than the preset confidence weight threshold, and determines the mapping relationship between facial features and corresponding terminal features based on the above spatiotemporal correlations.

[0184] As can be seen from the above steps, referring to Table 12 below, the corresponding mapping relationship set is as follows:

[0185] Table 12

[0186] Mapping relationship Confidence weights F(1)-IMSI(1) 2 F(4)-IMSI(3) 3 F(5)-IMSI(5) 4

[0187] Based on the above process, the collection time and location of key data are analyzed to determine the corresponding spatiotemporal correlation. Since the spatiotemporal correlation is inevitable and is not affected by other external factors, the mapping relationship obtained in this way is more accurate than the correlation obtained by deep fitting in related technologies. This avoids the problem of subsequent error transmission due to insufficient accuracy of the correlation between key data.

[0188] See Figure 11 The image shows a tracking method based on an electronic fence proposed in this application, comprising:

[0189] Step 111: Acquire the target facial image of the target object through the camera device, obtain the current target location information of the target object, and perform image processing on the target facial image to obtain the target facial features of the target object.

[0190] Specifically, in this embodiment of the application, the management platform obtains the target facial image of the target object and the target location information of the target object. Furthermore, the management platform can obtain the target location information of the target object through the deployment location of the camera device, and use a facial detection algorithm to obtain the target facial features of the target object by performing image processing on the target facial image.

[0191] For example, see Figure 12As shown, the management platform determines the current target location information of the target object as (104.00000, 36.00000) based on the deployment location of camera device x. Furthermore, the management platform obtains the target object's facial image through camera device x and uses a multi-task Cascaded Convolutional Network (MTCNN) algorithm to obtain the target object's facial feature information L = [L0, L1, ..., L...]. 128 ].

[0192] Step 112: Based on the preset mapping relationship set, determine the target terminal features corresponding to the target facial features.

[0193] Specifically, the above mapping relationship set includes: the association between the facial features of each candidate object and the terminal features of each candidate terminal, wherein each association is determined based on the facial feature acquisition location and facial feature acquisition time of the candidate object, and the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal.

[0194] Taking the mapping relationship set shown in Table 11 as an example, assume that the target facial feature information of the target object is L = [L0, L1, ..., L...]. 128 If the facial feature F(1) in the mapping relationship set is consistent with the target facial feature, then the target terminal feature corresponding to the target facial feature is determined to be IMSI(1) according to the mapping relationship F(1)-IMSI(1).

[0195] Step 113: When it is determined that the movement trajectory of the target object exceeds the preset electronic fence range based on the obtained target location information, the target object is tracked based on the target terminal characteristics.

[0196] Specifically, when the movement trajectory of the target object is determined to be outside the preset electronic fence range based on the obtained target location information, the GPS function of the target terminal is requested based on the characteristics of the target terminal to track the target object.

[0197] For example, see Figure 13 As shown, assuming the preset electronic fence range is: longitude coordinate range (104.50000, 105.00000), latitude coordinate range (36.50000, 37.00000); and the target location information is (104.00000, 36.00000), then based on the target location information, it is determined that the target object's movement trajectory exceeds the preset electronic fence range. Based on the obtained target terminal feature IMSI (1), the target terminal carried by the target object is tracked and queried, and the GPS function of the target terminal is requested, so as to track the target object.

[0198] See Figure 14 As shown, an embodiment of this application provides a tracking device based on an electronic fence, comprising: a data acquisition module 141, a determination module 142, and a tracking module 143, wherein:

[0199] The acquisition module 141 is used to acquire the target facial image of the target object through the camera device, acquire the current target position information of the target object, and perform image processing on the target facial image to obtain the target facial features of the target object.

[0200] Determining module 142: used to determine the target terminal features corresponding to the target facial features based on a preset mapping relationship set, wherein the mapping relationship set includes: the association relationship between the facial features of each candidate object and the terminal features of each candidate terminal, and each association relationship is determined based on the facial feature acquisition location and facial feature acquisition time of the candidate object, and the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal.

[0201] The tracking module 143 is used to track the target object based on the obtained target terminal characteristics when the target location information determines that the target object's movement trajectory exceeds the preset electronic fence range.

[0202] In an optional embodiment, before acquiring the target facial features of the target object through the camera device, the acquisition module 141 is further configured to:

[0203] Obtain the first candidate feature data set of each candidate object collected; wherein each first candidate feature data set includes at least: the facial features of the corresponding candidate object, the facial feature acquisition time, and the facial feature acquisition location.

[0204] Obtain the second candidate feature data set of each candidate terminal collected; wherein each second candidate feature data set includes at least: the terminal feature of the corresponding candidate terminal, the terminal feature collection time, and the terminal feature collection location.

[0205] Based on the acquisition time and location of each facial feature in the first candidate feature data set, and the acquisition time and location of each terminal feature in the second candidate feature data set, the spatiotemporal correlation between each acquired facial feature and the corresponding terminal feature is determined.

[0206] The spatiotemporal correlation between each facial feature and its corresponding terminal feature is input into a preset correlation model to obtain the mapping relationship between each facial feature and its corresponding terminal feature.

[0207] In an optional embodiment, when acquiring the first set of candidate feature data for each candidate object, the acquisition module 141 is specifically used for:

[0208] Using a time window overlay algorithm, based on a preset time range, the first candidate feature data set of each candidate object is selected from the first candidate feature data obtained by feature collection for each object.

[0209] In an optional embodiment, when acquiring the second candidate feature data set of each candidate terminal, the acquisition module 141 is specifically used for:

[0210] Using a time window overlay algorithm, based on a preset time range, the second candidate feature data set of each candidate terminal is selected from the second candidate feature data obtained by feature collection for each terminal.

[0211] In an optional embodiment, when determining the spatiotemporal correlation between each collected facial feature and its corresponding terminal feature based on the facial feature acquisition time and location in the obtained first candidate feature data set, and the terminal feature acquisition time and location in the obtained second candidate feature data set, the acquisition module 141 is specifically used for:

[0212] Based on the acquisition time of each facial feature in the first candidate feature data set and the acquisition time of each terminal feature in the second candidate feature data set, the temporal correlation between each facial feature and its corresponding terminal feature is determined.

[0213] Based on the facial feature acquisition locations in the first candidate feature data set and the terminal feature acquisition locations in the second candidate feature data set, the spatial association between each facial feature and its corresponding terminal feature is determined.

[0214] Based on the obtained temporal and spatial correlations, the spatiotemporal correlations between each collected facial feature and the corresponding terminal feature are determined.

[0215] In an optional embodiment, when determining the temporal correlation between each facial feature and its corresponding terminal feature based on the acquisition time of each facial feature in the obtained first candidate feature data set and the acquisition time of each terminal feature in the obtained second candidate feature data set, the acquisition module 141 is specifically used for:

[0216] For each facial feature acquisition time and the corresponding terminal feature acquisition time, perform the following operations respectively:

[0217] Determine the time difference between a facial feature acquisition time and the corresponding terminal feature acquisition time.

[0218] If the obtained time difference is less than the preset time difference threshold, the temporal correlation between facial features and corresponding terminal features is determined based on the time difference.

[0219] If the obtained time difference is not less than the preset time difference threshold, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the second candidate feature data set.

[0220] In an optional embodiment, when determining the spatial association between each facial feature and its corresponding terminal feature based on the facial feature acquisition locations in the obtained first candidate feature data set and the terminal feature acquisition locations in the obtained second candidate feature data set, the acquisition module 141 is specifically used for:

[0221] For each facial feature acquisition location and the corresponding terminal feature acquisition location, perform the following operations respectively:

[0222] Obtain the physical distance between a facial feature acquisition location and the corresponding terminal feature acquisition location.

[0223] If the obtained physical distance is less than the preset distance threshold, the spatial relationship between facial features and corresponding terminal features is determined based on the physical distance.

[0224] If the obtained physical distance is not less than the preset standard distance, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the obtained second candidate feature data set.

[0225] In an optional embodiment, when inputting the spatiotemporal correlation between the obtained facial features and corresponding terminal features into a preset correlation model to obtain the mapping relationship between each facial feature and its corresponding terminal feature, the acquisition module 141 is specifically used for:

[0226] For each facial feature and its corresponding terminal feature, perform the following operations respectively:

[0227] Input the spatiotemporal correlation between a facial feature and its corresponding terminal feature into a preset correlation model to obtain the confidence weights corresponding to the spatiotemporal correlation.

[0228] If the obtained confidence weight is less than the preset confidence weight threshold, then the corresponding facial features are removed from the obtained first candidate feature data set, and the corresponding terminal features are removed from the obtained second candidate feature data set.

[0229] If the obtained confidence weight is not less than the preset confidence weight threshold, then the mapping relationship between facial features and corresponding terminal features is determined based on the obtained spatiotemporal correlation.

[0230] Based on the same inventive concept as the embodiments described above, this application also provides an electronic device that can be used for tracking based on electronic fences. In one embodiment, the electronic device can be a server, a terminal device, or other electronic equipment. In this embodiment, the structure of the electronic device can be as follows: Figure 15 As shown, it includes a memory 151, a communication interface 153, and one or more processors 152.

[0231] The memory 151 is used to store computer programs executed by the processor 152. The memory 151 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0232] Memory 151 may be volatile memory, such as random-access memory (RAM); memory 151 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 151 may be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 151 may be a combination of the above-described memories.

[0233] Processor 152 may include one or more central processing units (CPUs) or digital processing units, etc. Processor 152 is used to implement the above-described electronic fence-based tracking method when it calls the computer program stored in memory 151.

[0234] Communication interface 153 is used to communicate with terminal devices and other servers.

[0235] This application embodiment does not limit the specific connection medium between the memory 151, the communication interface 153, and the processor 152. This application embodiment... Figure 15 The memory 151 and the processor 152 are connected via a bus 154, and the bus 154 is in Figure 15The connections between other components are shown in thick lines only and are not intended to be limiting. The bus 154 can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 15 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0236] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform any of the electronic fence-based tracking methods described above. The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0237] In this embodiment, image processing is performed on the acquired target facial image to obtain the target facial features of the target object. Based on the target facial features and a preset mapping relationship set, the corresponding target terminal features are determined. Since the mapping relationship set contains the correlation between the facial features of each candidate object and the terminal features of each candidate terminal, and each mapping relationship is determined only by the facial feature acquisition location and facial feature acquisition time of the candidate object, as well as the terminal feature acquisition location and terminal feature acquisition time of the corresponding candidate terminal, the correlation between each mapping relationship in the mapping relationship set is higher. That is, the correlation between the obtained target facial features and target terminal features is higher, thereby further improving the tracking accuracy.

[0238] Furthermore, as can be seen from the above process, this method only needs to match the obtained target facial features with a small number of facial features stored in the mapping relationship set to directly and accurately obtain the target terminal features, thereby effectively improving the correlation efficiency of key data during the tracking process, making the obtained action trajectory more complete, and providing more time for action to discover the target object's location.

[0239] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0240] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0241] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0242] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0243] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A tracking method based on electronic fences, characterized in that, include: The target facial image of the target object is acquired by a camera device, and the target facial image is processed to obtain the target facial features of the target object. Based on a preset set of mapping relationships, the target terminal features corresponding to the target facial features are determined; wherein, the set of mapping relationships is determined in the following manner: Obtain a first set of candidate feature data for each candidate object, and a second set of candidate feature data for each candidate terminal; wherein each first set of candidate feature data includes at least the facial features, facial feature acquisition time, and facial feature acquisition location of the corresponding candidate object; each second set of candidate feature data includes at least the terminal features, terminal feature acquisition time, and terminal feature acquisition location of the corresponding candidate terminal. Based on the facial feature acquisition time and location in the first candidate feature data set, and the terminal feature acquisition time and location in the second candidate feature data set, the spatiotemporal relationship between each candidate object and each candidate terminal is determined. Based on the set rules, the spatiotemporal correlations of each group are optimized to obtain the optimized spatiotemporal correlations of each group; wherein, for any group of spatiotemporal correlations, the set rules are used to indicate the removal of candidate objects corresponding to facial features and candidate terminals corresponding to terminal features in the spatiotemporal correlations where the number of occurrences in the target area within a set time range exceeds a threshold; the target area is determined based on the location of at least one mobile base station set within the area; Calculate the confidence weight of each optimized spatiotemporal correlation relationship, and select the target facial features and target terminal features indicated by the optimized spatiotemporal correlation relationships with confidence weights greater than the preset confidence weight threshold to form the mapping relationship set.

2. The method as described in claim 1, characterized in that, The step of determining the spatiotemporal correlation between each facial feature and its corresponding terminal feature based on the facial feature acquisition time and location in the first candidate feature data set, and the terminal feature acquisition time and location in the second candidate feature data set, includes: Based on the acquisition time of each facial feature in the first candidate feature data set and the acquisition time of each terminal feature in the second candidate feature data set, the temporal correlation between each facial feature and the corresponding terminal feature is determined respectively. Based on the facial feature acquisition locations in the first candidate feature data set and the terminal feature acquisition locations in the second candidate feature data set, the spatial association between each facial feature and its corresponding terminal feature is determined. Based on the temporal and spatial correlations, the spatiotemporal correlations between each collected facial feature and the corresponding terminal feature are determined.

3. The method as described in claim 2, characterized in that, The step of determining the temporal correlation between each facial feature and its corresponding terminal feature based on the acquisition time of each facial feature in the first candidate feature data set and the acquisition time of each terminal feature in the second candidate feature data set includes: For each facial feature acquisition time and the corresponding terminal feature acquisition time, perform the following operations respectively: Determine the time difference between a facial feature acquisition time and the corresponding terminal feature acquisition time; If the time difference is less than a preset time difference threshold, then based on the time difference, the temporal correlation between the facial feature and the corresponding terminal feature is determined; If the time difference is not less than a preset time difference threshold, then the corresponding facial features are removed from the first candidate feature data set, and the corresponding terminal features are removed from the second candidate feature data set.

4. The method as described in claim 2, characterized in that, The step of determining the spatial association between each facial feature and its corresponding terminal feature based on the facial feature acquisition locations in the first candidate feature data set and the terminal feature acquisition locations in the second candidate feature data set includes: For each facial feature acquisition location and the corresponding terminal feature acquisition location, perform the following operations respectively: Obtain the physical distance between a facial feature acquisition location and the corresponding terminal feature acquisition location; If the physical distance is less than a preset distance threshold, then based on the physical distance, the spatial association between the facial feature and the corresponding terminal feature is determined; If the physical distance is not less than a preset distance threshold, then the corresponding facial features are removed from the first candidate feature data set, and the corresponding terminal features are removed from the second candidate feature data set.

5. The method according to any one of claims 1 to 4, characterized in that, The calculation of the confidence weight of each optimized spatiotemporal correlation relationship, and the selection of target facial features and target terminal features indicated by optimized spatiotemporal correlation relationships with confidence weights greater than a preset confidence weight threshold to form the mapping relationship set, including: Each optimized spatiotemporal correlation is input into a preset correlation model to obtain the confidence weight corresponding to the spatiotemporal correlation. If the confidence weight is less than the preset confidence weight threshold, then the corresponding facial features and the corresponding terminal features are removed. If the confidence weight is not less than the preset confidence weight threshold, then based on the spatiotemporal correlation, the mapping relationship between the facial feature and the corresponding terminal feature is determined; The multiple mapping relationships are determined to constitute the mapping relationship set.

6. A tracking device based on an electronic fence, characterized in that, include: The acquisition module is used to acquire a target facial image of a target object through a camera device, and to perform image processing on the target facial image to obtain the target facial features of the target object; The determination module is used to determine the target terminal features corresponding to the target facial features based on a preset set of mapping relationships; The device is used to determine the mapping relationship set in the following manner: Obtain a first set of candidate feature data for each candidate object, and a second set of candidate feature data for each candidate terminal; wherein each first set of candidate feature data includes at least the facial features, facial feature acquisition time, and facial feature acquisition location of the corresponding candidate object; each second set of candidate feature data includes at least the terminal features, terminal feature acquisition time, and terminal feature acquisition location of the corresponding candidate terminal. Based on the facial feature acquisition time and location in the first candidate feature data set, and the terminal feature acquisition time and location in the second candidate feature data set, the spatiotemporal relationship between each candidate object and each candidate terminal is determined; wherein, one candidate object is associated with at least one candidate terminal, and one candidate terminal is associated with at least one candidate object. Based on the set rules, the spatiotemporal correlations of each group are optimized to obtain the optimized spatiotemporal correlations of each group; wherein, for any group of spatiotemporal correlations, the set rules are used to indicate the removal of candidate objects corresponding to facial features and candidate terminals corresponding to terminal features in the spatiotemporal correlations where the number of occurrences in the target area within a set time range exceeds a threshold; the target area is determined based on the location of at least one mobile base station set within the area; Calculate the confidence weight of each optimized spatiotemporal correlation relationship, and select the target facial features and target terminal features indicated by the optimized spatiotemporal correlation relationships with confidence weights greater than the preset confidence weight threshold to form the mapping relationship set.

7. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the tracking method based on an electronic fence as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-5.

Images