A companion determination method, electronic device, and storage medium

Abstract

Embodiments of the present application disclose a same-person determination method, an electronic device and a storage medium. The method comprises: determining a time range, a start area and an end area for pedestrian data screening according to trajectory information of a target person; obtaining trajectory information of a plurality of pedestrians matched according to the time range, the start area and the end area, and aggregating to obtain a plurality of trajectory clusters; selecting at least one cluster meeting a first clustering standard from the plurality of trajectory clusters; determining a pedestrian who is not a target person among pedestrians corresponding to trajectory objects in the selected cluster as a first type of pedestrian; and determining the first type of pedestrian as a same-person of the target person. The same-person determination scheme provided by the present disclosure can effectively identify other same-persons of the target person, and further improve the big data analysis function of the security system.

Description

Technical Field

[0001] This invention relates to, but is not limited to, the field of security monitoring, and particularly to a method for identifying companions, an electronic device, and a storage medium. Background Technology

[0002] Video surveillance is a crucial component of security systems. With advancements in video detection and recognition technologies, as well as the development of big data analytics, pedestrian data collected by networked monitoring devices can be used to create tracking strategies for individuals. Further enhancing the big data analytics capabilities of security systems involves identifying individuals exhibiting characteristics of organized groups based on known targets. Summary of the Invention

[0003] This disclosure provides a method, electronic device, and storage medium for identifying companions. Based on the trajectory information of the target person, it determines matching pedestrian trajectory information and aggregates trajectory objects, which can effectively identify other companions of the target person and further improve the big data analysis function of the security system.

[0004] On the one hand, embodiments of this disclosure provide a method for determining fellow travelers, including:

[0005] Based on the trajectory information of the target person, determine the time range, starting area, and ending area for pedestrian data filtering;

[0006] Based on the time range, starting area, and ending area, obtain the trajectory information of multiple matching pedestrians;

[0007] Multiple trajectory clusters are obtained by aggregating trajectory objects based on the trajectory information of the multiple pedestrians; at least one cluster that meets the first clustering criterion is selected from the multiple trajectory clusters; pedestrians who are not the target person among the trajectory objects contained in the selected at least one cluster are determined to be the first type of pedestrians;

[0008] The first type of pedestrians are identified as companions of the target person.

[0009] On the other hand, embodiments of this disclosure also provide an electronic device, including:

[0010] One or more processors;

[0011] Storage device for storing one or more programs.

[0012] When the one or more programs are executed by the one or more processors, the one or more processors implement the peer identification method as described in any embodiment of this disclosure.

[0013] On the other hand, embodiments of this disclosure also provide a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the co-op determination method as described in any embodiment of this disclosure.

[0014] After reading and understanding the accompanying diagrams and detailed descriptions, the other aspects can be understood. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a method for determining fellow travelers provided in an embodiment of the present invention;

[0017] Figure 2 This is a flowchart of a method for determining fellow travelers provided in an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram of trajectory weight discrimination and identification of abnormal trajectories provided in an embodiment of the present invention;

[0019] Figure 4 This is a flowchart of another method for determining fellow personnel provided in an embodiment of the present invention.

[0020] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0023] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0024] In this invention, unless otherwise explicitly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0026] This disclosure provides a method for determining companions, such as... Figure 1 As shown, it includes:

[0027] Step 110: Based on the trajectory information of the target person, determine the time range, starting area, and ending area for pedestrian data filtering;

[0028] Step 120: Based on the time range, starting area, and ending area, obtain the trajectory information of multiple matching pedestrians;

[0029] Step 130: Aggregate trajectory objects based on the trajectory information of the multiple pedestrians to obtain multiple trajectory clusters; select at least one cluster that meets the first clustering criterion from the multiple trajectory clusters; determine that pedestrians who are not the target person among the trajectory objects contained in the selected at least one cluster are the first type of pedestrians;

[0030] Step 140: Identify the first type of pedestrians as companions of the target person.

[0031] In some exemplary embodiments, such as Figure 2 As shown, step 140 includes:

[0032] Step 140-2: Based on the time period of stay of the first type of pedestrians after arriving at the destination area and the time period of stay of the target person after arriving at the destination area, determine the overlap time period of each first type of pedestrian and the target person in the destination area, and select the first type of pedestrians whose overlap time period meets the time overlap standard as the companions of the target person.

[0033] As can be seen, in some exemplary embodiments, after aggregating trajectory objects based on trajectory information, pedestrians who are not the target person in the clusters that meet the first clustering criteria are considered as companions of the target person. In other exemplary embodiments, time overlap filtering of the destination area can be further performed to further narrow down the range of companions, thereby reducing the amount of data for subsequent data processing of companions determined based on the embodiments of this disclosure. In specific embodiments, it can be selected whether to further perform time overlap filtering based on the behavioral characteristics of the target person and companions and / or the number of the first type of pedestrians determined in step 130.

[0034] In some exemplary embodiments, the trajectory information refers to the collection of information from the trajectory points traversed by the pedestrian from the starting area to the ending area, and the collection information for each trajectory point includes: location and time.

[0035] In some exemplary embodiments, the location of a point includes longitude and latitude.

[0036] For example, a pedestrian's trajectory includes m trajectory points, and its trajectory information tr i Recorded as:

[0037] tr i ={{lat1,lng1,t1},{lat2,lng2,t2},…,{lat m ,lng m ,t m}};

[0038] Where lat is latitude, lng is longitude, and t is time. tr i The above description is also known as the trajectory point model description. In some exemplary embodiments, during clustering, each trajectory information corresponds to a trajectory object, also denoted as tr. i .

[0039] Alternatively, the location of the trajectory points can be represented in other ways.

[0040] It should be noted that the data collection devices / systems at the trajectory points include: cameras, RFID data collection devices, and / or WiFi hotspots, etc., which have the function of collecting information on pedestrians and / or devices carried by pedestrians. The specific data collection scheme is not discussed in this embodiment.

[0041] In some exemplary embodiments, step 120, based on pre-established pedestrian profile information, can group multiple categories of collected data for the same pedestrian obtained by various collection devices / systems into a single pedestrian, forming complete trajectory information. The specific aspects of grouping multiple categories of collected data into a single pedestrian are implemented according to relevant schemes, and will not be discussed in detail in this application.

[0042] In some exemplary embodiments, the pedestrian profile information may be obtained in advance from one or more external information systems, and specific aspects are not discussed in the embodiments of this application.

[0043] In some exemplary embodiments, the pedestrian profile information includes one or more of the following:

[0044] Basic population information, device information of the associated mobile device, facial information, and telephone information.

[0045] The basic population information includes one or more of the following:

[0046] Name, gender, ID number, place of residence, and workplace.

[0047] The device information belonging to the mobile device includes: the MAC address information of the mobile device.

[0048] It should be noted that pedestrian profile information can be expanded as needed, and is not limited to the aspects shown in the examples above.

[0049] In some exemplary embodiments, step 130, determining whether a trajectory cluster conforms to a first clustering criterion, includes:

[0050] The proportion of trajectory objects included in a trajectory cluster relative to the total number of trajectory objects included in all trajectory clusters is calculated and denoted as the proportion of trajectory objects in a trajectory cluster.

[0051] If the proportion of trajectory objects is less than a set first threshold, the trajectory cluster is determined to meet the first clustering criterion; if the proportion of trajectory objects is greater than or equal to the set first threshold, the trajectory cluster is determined to not meet the first clustering criterion.

[0052] As can be seen, in some exemplary embodiments, trajectory clusters that meet the first clustering criterion are those with a small proportion of trajectory objects. This means that these trajectory clusters contain a relatively small number of trajectory objects, indicating that the pedestrians corresponding to these trajectory objects have chosen non-common routes. When the target is a suspect and their companions are aware of counter-tracking, they may deliberately choose different or unconventional routes. Therefore, in related embodiments, when tracking down and pursuing a suspect's companions, these trajectory clusters that meet the first clustering criterion are also referred to as abnormal trajectory clusters.

[0053] In some exemplary embodiments, step 110 includes:

[0054] Based on the starting point in the target person's trajectory information, determine the starting area SA, and based on the ending point in the trajectory information, determine the ending area EA.

[0055] Based on the time ST in the starting point acquisition information, the time DT in the ending point acquisition information, and the set redundancy duration T1, the time range is determined to be from ST-T1 to DT+T1, denoted as [ST-T1, DT+T1].

[0056] It should be noted that determining the starting region SA based on the starting point of the target person's trajectory information and the ending region EA based on the ending point can be done manually or according to relevant rules based on relevant map data and data from the acquisition device / system. This is not limited to any specific aspect. The larger the redundancy duration T1, the larger the corresponding time window, and the more trajectory information to be acquired and processed. It can be set according to the specific application scenario and is not limited to specific values.

[0057] In some exemplary embodiments, step 120 involves obtaining the trajectory information of multiple matching pedestrians based on the time range, the starting area, and the ending area. Specifically, based on the pedestrian information collected by the monitoring system at various collection points, the trajectory information of multiple pedestrians collected from the starting area to the ending area within the time range is matched. The trajectory information of these multiple pedestrians is also referred to as a trajectory information set, denoted as:

[0058] OT = {tr1,tr2,tr3,…,tr k}

[0059] The trajectory of a pedestrian includes m trajectory points, and its trajectory information tr i Recorded as:

[0060] tr i ={{lat1,lng1,t1},{lat2,lng2,t2},…,{lat m ,lng m ,tm}};1≤i≤k

[0061] The number of trajectory points m included in the trajectory information of different pedestrians can be the same or different.

[0062] In some exemplary embodiments, the trajectory information of multiple pedestrians obtained in step 120 includes the trajectory information of the target person and the trajectory information of other non-target persons. The solution provided by this disclosure embodiment is to find other persons traveling with the target person from these pedestrians through clustering and filtering. Especially when the target person is a suspect, the solution provided by this disclosure embodiment can find other accomplices with spatiotemporal correlation.

[0063] In some exemplary embodiments, step 120 includes:

[0064] Based on the determined time range, starting region, and ending region, the locations of all matching collection points are obtained. These locations are then aggregated according to a preset proximity threshold θ, forming the location of the trajectory points. In other words, collection points with a distance less than the threshold θ are grouped together as a single location.

[0065] As can be seen, due to the diversity of data sources, the latitude and longitude values ​​of the same location may differ slightly in different categories of data. To reduce the complexity of data calculation and analysis, it is necessary to aggregate trajectory points for nearby data. The distance between adjacent data collection points is calculated. For two points whose distance is less than a preset proximity distance threshold θ (e.g., 5m), the latitude and longitude of the points are set to the same value (the value of any object trajectory point can be used). This method ensures the consistency of latitude and longitude at the same trajectory point for all trajectories.

[0066] In some exemplary embodiments, step 130 involves aggregating trajectory objects based on the trajectory information of the multiple pedestrians to obtain multiple trajectory clusters, including:

[0067] The trajectory objects corresponding to the trajectory information of the multiple pedestrians are aggregated according to the trajectory similarity between the trajectories to obtain the multiple trajectory clusters;

[0068] The trajectory similarity between the trajectories is calculated based on the spatial similarity and temporal similarity between the trajectories.

[0069] In some exemplary embodiments, the OWD (One Way Distance) algorithm is used as the trajectory space similarity calculation method, as follows:

[0070]

[0071]

[0072] d(p,tr i ) represents the trajectory tr base GPS point p to tr i The distance from point p to tr i The minimum Euclidean ensemble distance between all points on the trajectory. d(p,tr) i The calculation is as follows:

[0073]

[0074]

[0075] lat p , lng p lat q , lng q Let represent the latitude and longitude of points p and q, respectively. According to the equation, d... owd (tr base ,tr i The smaller the value, the higher the spatial similarity between the two trajectories.

[0076] Alternatively, other algorithms can be used to calculate spatial similarity, such as multi-line location distance algorithm, Fraser distance algorithm, etc., and are not limited to the aspects of the embodiments of this disclosure.

[0077] Those skilled in the art will understand that the above algorithm can be used to calculate the spatial similarity between each pair of multiple pedestrian trajectories.

[0078] In some exemplary embodiments, the temporal similarity between trajectories is calculated according to the following method:

[0079] For each trajectory information, a trajectory segment model (TRLS) is established for the trajectory object, also known as a trajectory segment model description:

[0080] trls i =trans(tr i )

[0081] trls i ={{sp1,ep1,tc1},{sp2,ep2,tc2}…,{sp m-1 ,ep m-1 ,tc m-1}};1≤i≤k;

[0082] The difference between the trajectory segment model description and the trajectory point model description is that the segment model represents the description of the trajectory segment formed by each two spatiotemporal data acquisitions of the object. sp represents the starting position of the segment, ep represents the ending position of the segment, and tc represents the time taken to pass through the segment (i.e., the trajectory segment with the starting point sp and the ending point ep, also known as the road segment).

[0083] Based on the above trajectory segment description, it can be known that if two pedestrians pass through the same road segment, the sp and ep values ​​of this same road segment will be the same in the trajectory segment descriptions of the two pedestrians.

[0084] Calculate the time similarity between trajectories using the following formula:

[0085]

[0086] ∑tcls1 same With ∑tcls2 same It is the total time taken by pedestrians on two tracks, 1 and 2, to traverse the same road segment. sumtime(trls1) represents the total time taken for track trls1, and sumtime(trls2) represents the total time taken for track trls2.

[0087] The calculation results show that 0 ≤ d time (trls1,trls2)≤1, the larger the value, the greater the time similarity between trajectory 1 and 2.

[0088] In some exemplary embodiments, the trajectory similarity between trajectories is calculated based on the spatial similarity and temporal similarity between trajectories, including:

[0089] d(tr1,tr2)=d owd (tr1,tr2)*(1-d time (trls1,trls2))

[0090] Where d(tr1,tr2) is the similarity between trajectories 1 and 2, d owd (tr1, tr2) represents the spatial similarity between trajectories 1 and 2, d time (trls1, trls2) represents the temporal similarity between trajectories 1 and 2. It can be seen that the smaller the similarity d(tr1, tr2), the higher the similarity between the trajectories.

[0091] In some exemplary embodiments, step 130, which aggregates the trajectory objects corresponding to the trajectory information of the multiple pedestrians according to the trajectory similarity between trajectories, includes:

[0092] Step 130110: Calculate the trajectory weight of each pedestrian trajectory according to the set road weight;

[0093] Step 130120: Use the plurality of trajectory objects as the initial atomic trajectory objects;

[0094] Step 130130: Perform the following aggregation and filtering steps for all atomic trajectory objects:

[0095] The atomic trajectory objects are aggregated according to the trajectory similarity between trajectories to obtain at least one trajectory cluster; the trajectory cluster includes: multi-object trajectory cluster and / or first single-object trajectory cluster;

[0096] Based on the trajectory weights of the trajectory objects included in each of the object trajectory clusters, trajectory objects whose trajectory weights meet the trajectory offset criteria are removed from the multi-object trajectory clusters to which they belong and treated as a second single-object trajectory cluster.

[0097] The trajectory objects in the first and second single-object trajectory clusters are used as atomic trajectory objects;

[0098] Continue executing the aggregation and filtering steps 130130 until all trajectory clusters meet the clustering convergence criteria.

[0099] In some exemplary embodiments, the road weight refers to the weight of each road segment set according to the number of people passing through each road segment in the pedestrian trajectory. The number of people passing through each road segment can be determined based on statistical data from historically collected pedestrian data, or based on pedestrian trajectory information obtained in step 120, or based on data from other systems, and is not limited to any specific form.

[0100] In some exemplary embodiments, the trajectory weight of a pedestrian trajectory is the sum of the weights of all (road segments) included in the pedestrian trajectory.

[0101] For example, a pedestrian's trajectory is

[0102] tr i ={{lat1,lng1,t1},{lat2,lng2,t2},…,{lat m ,lng m ,t m}};1≤i≤k

[0103] The trajectory includes: segment 1 (lat1, lng1) - (lat2, lng2), segment 2 (lat2, lng2) - (lat3, lng3), ... segment m-1 (lat1, lng1) - (lat2, lng2), segment m-1 (lat2, lng2), ... m-1 ,lng m-1 )-(lat m ,lng m ).

[0104] The weights of each road segment are: v1, v2, ... v m-1 Then the trajectory tr i trajectory weight w i = v1 + v2 + ... + v m-1 .

[0105] Accordingly, the trajectory weights of multiple pedestrian trajectories are calculated separately, resulting in the following weighted trajectory object OTW:

[0106] OTW={{tr1,w1},{tr2,w2},{tr3,w3},…,{tr k ,w k}}

[0107] Where tr is the trajectory object and w is the corresponding trajectory weight.

[0108] In some exemplary embodiments, the trajectory weights are determined to conform to the trajectory offset criteria according to the following method:

[0109] The weight deviation corresponding to the trajectory weight is determined based on the average of the trajectory weights and the clusters to which they belong.

[0110] If the weight deviation is greater than a preset deviation threshold, the trajectory weight is determined to conform to the trajectory offset standard.

[0111] If the weight deviation is less than or equal to a preset deviation threshold, it is determined that the trajectory weight does not meet the trajectory offset standard.

[0112] The mean trajectory weight is the average of the trajectory weights of all pedestrian trajectories in the assigned cluster.

[0113] In some exemplary embodiments, taking the trajectory object tr1 in cluster C'1 as an example, its weight bias E(tr1) is calculated as follows:

[0114]

[0115] weight(tr h ) represents the trajectory tr h The trajectory weights are obtained from the OTW determined in step 130110, where number(C'1) represents the number of trajectory objects in cluster C'1. It can be seen that the further an object deviates from the average weight value, the greater its deviation value.

[0116] In some exemplary embodiments, all trajectory clusters are determined to satisfy the clustering convergence criterion according to the following method:

[0117] If the total number of trajectory clusters remains constant, and the trajectory objects contained in each trajectory cluster remain constant, then the clustering convergence criterion is satisfied.

[0118] If the total number of trajectory clusters changes, or if the trajectory objects contained in any trajectory cluster change, the clustering convergence criterion is determined not to be met.

[0119] Those skilled in the art will understand that the unchanged total number of trajectory clusters indicates that the total number of clusters obtained by aggregation tends to be stable; the unchanged number of trajectory objects contained in the trajectory clusters indicates that the number of objects contained is unchanged, and the objects themselves are also unchanged. The above clustering and filtering steps 130 are executed iteratively until all trajectory clusters meet the clustering convergence criteria, indicating that the above steps are executed iteratively until the clustering results obtained by the clustering remain stable.

[0120] In some exemplary embodiments, step 130, which aggregates the trajectory objects corresponding to the trajectory information of the multiple pedestrians according to the trajectory similarity between trajectories, includes:

[0121] Step 130210: Use each trajectory object as an initial atomic trajectory object (atomic state);

[0122] Perform the following steps on all atomic trajectory objects:

[0123] Step 130220: Set the trajectory similarity threshold ε. Initialize each atomic trajectory object into an independent cluster (aggregate region). First, take the first atomic trajectory object tr1 and calculate the trajectory similarity value between tr1 and other trajectory objects. If tr... i When the trajectory similarity value with tr1 is the lowest (the lower the value, the more similar the trajectory), and is below the threshold ε, tr1 and tr... i Aggregates into one cluster. Then, using the second atomic object (the unaggregated single object), similarity is calculated with other trajectory objects. If it matches the trajectory tr... m The trajectory similarity value is the lowest and below the threshold ε. If tr m If the current state is atomic, then the two trajectories merge into one class, if tr m If a data point is already a member of a cluster, it needs to be further evaluated for trajectory similarity with other members of that cluster. If the trajectory similarity value with other members is also lower than ε, then tr... m Join the cluster; if the similarity value with one or more members of the cluster is higher than ε, then keep tr. m The object is an atomic trajectory object (also called an atomic node). This process continues until all atomic trajectory objects have been computed, and no remaining atomic trajectory objects can be aggregated. The initial aggregation result is then obtained:

[0124] C' = {C'1,C'2,C'3,…,C'} p};0 <p<k′

[0125] Among them, C' iIt can be either multi-object trajectory clustering or single-object trajectory clustering. Multi-object trajectory clustering is a cluster (aggregate group) that contains multiple trajectory objects, while single-object trajectory clustering is a cluster (aggregate group) that contains only one trajectory object. Here, it is referred to as the first single-object trajectory clustering.

[0126] Step 130230: Based on the weight value of each trajectory object in OTW, extract the trajectory whose trajectory weight deviates too much from that of other objects in each multi-object trajectory cluster.

[0127] Taking a cluster as an example, with a preset deviation threshold of ξ, steps 130230 include:

[0128] Step 130230-1: Calculate the weight bias of each trajectory object in the cluster using the following method:

[0129]

[0130] E(tr o ) is the trajectory object tr o In the cluster C' to which it belongs j The weight bias in the weight(tr) o ) represents the trajectory object tr o Trajectory weights, weight(tr h ) represents the trajectory object tr h The trajectory weights, number(C' j ) represents clustering C' j The number of trajectory objects. It can be seen that the further an object deviates from the average weight value, the greater the deviation value obtained;

[0131] 130230-2, if the weight deviation of the trajectory object is greater than the preset deviation threshold ξ, then the trajectory object tr o Remove it from its cluster and remove the trajectory object tr o As a single-object trajectory cluster, it is denoted as the second single-object trajectory cluster;

[0132] For all multi-object trajectory clusters obtained in step 130220, execute steps 130230-1 and 130230-2 one by one to complete the filtering of deviating trajectory objects in all multi-object trajectory clusters.

[0133] Take the trajectory objects from the first single-object trajectory cluster obtained in step 130220 and the trajectory objects from the second single-object trajectory cluster obtained in step 130230 as atomic trajectory objects, and repeat steps 130220 and 130230 until the clustering convergence criterion is met.

[0134] In some exemplary embodiments, such as Figure 2As shown, the solid line trajectory is the trajectory of the normal trajectory object X, and the dashed line trajectory is the trajectory of the abnormal trajectory object Y (trajectory points 1, 4, and 5 are separated for ease of presentation, but they are actually the same after aggregation through neighboring points). The dashed line is the actual trajectory of object Y from point 1 to point 4. According to the analysis above, the spatial similarity between X and Y is very high (d Euclid (X,Y) is very low), and Y can also maintain high time similarity through acceleration. Therefore, trajectory weights are added for secondary discrimination to identify such abnormal trajectories Y. As can be seen, in step 130230, abnormal trajectories in multi-object clustering are filtered according to the weight deviation of trajectory objects to form a second single-object trajectory cluster.

[0135] In some exemplary embodiments, step 130, which involves aggregating trajectory objects based on the trajectory information of the multiple pedestrians to obtain multiple trajectory clusters, further includes:

[0136] The obtained trajectory clusters are sorted in descending order according to the number of trajectory objects contained in each cluster, resulting in multiple sorted trajectory clusters C. The data in each Ci is a set of trajectory objects.

[0137] C = {C1, C2, C3, ..., Cp} 0 <p<k。

[0138] In some exemplary embodiments, step 130, determining whether a trajectory cluster conforms to a first clustering criterion, includes:

[0139] The proportion of trajectory objects included in a trajectory cluster relative to the total number of trajectory objects included in all trajectory clusters is calculated and denoted as the proportion of trajectory objects in a trajectory cluster.

[0140] If the proportion of trajectory objects is less than a set first threshold, the trajectory cluster is determined to meet the first clustering criterion; if the proportion of trajectory objects is greater than or equal to the set first threshold, the trajectory cluster is determined to not meet the first clustering criterion.

[0141] In some exemplary embodiments, the first threshold is 10%, that is, the proportion of the number of trajectory objects in each trajectory cluster relative to the total number of pedestrian trajectory objects is determined and denoted as the trajectory object proportion. Trajectory clusters with a trajectory object proportion of less than 10% are selected. Accordingly, pedestrians who are not the target person among the trajectory objects in these trajectory clusters are identified as first-category pedestrians to further confirm whether they are any of the target's companions. In some exemplary embodiments, these selected trajectory clusters are also called abnormal trajectory clusters, and the trajectory objects included in them are considered abnormal trajectories.

[0142] In some exemplary embodiments, based on the above trajectory clustering ranking result C, the trajectories whose proportion of selected trajectory objects is less than the first threshold are clustered as: Cq,…Cp (q<=p).

[0143] In some exemplary embodiments, if the trajectory object corresponding to the trajectory information of the target person is included in a trajectory cluster in which the proportion of selected trajectory objects is less than a first threshold, it indicates that the trajectory of the target person is also an abnormal trajectory. In this case, it can be preliminarily determined that the target person, like the first type of pedestrian, has anti-tracking awareness and has chosen an abnormal trajectory to reach the destination area from the starting area.

[0144] In some exemplary embodiments, the time period of stay of the first type of pedestrian and the target person in step 140-2 when they arrive at the destination area can be determined based on the stay information collected by the collection device in the destination area.

[0145] In some exemplary embodiments, the endpoint area includes multiple collection points. The collection point that first arrives at the endpoint area is denoted as the entry point, and the collection point that leaves the endpoint area is denoted as the exit point. The recording of the dwell information of the first type of pedestrian or target person in the endpoint area is: [{ST,POINT1},{ET,POINT2}], where ST represents the entry time, POINT1 represents the entry point location, ET represents the exit time, and POINT2 represents the exit point location. It should be noted that different pedestrians may have the same or different entry points in the endpoint area, and their exit points may also be the same or different.

[0146] In some exemplary embodiments, in step 140-2, for each first type of pedestrian, the overlap time between the first type of pedestrian and the target person in the destination area can be determined based on the entry time and exit time in their dwell information and the entry time and exit time in the target person's dwell information.

[0147] In some exemplary embodiments, step 140-2, determining whether the overlapping time period meets the time overlap criterion, includes:

[0148] Calculate the proportion of the overlapping time period relative to the dwell time of the corresponding first type of pedestrians after arriving at the destination area, and record it as the overlap time ratio of the first type of pedestrians.

[0149] Calculate the proportion of the overlapping time period relative to the time the target person stays after arriving at the destination area, and record it as the overlap time ratio of the target person;

[0150] If the overlap time ratio of the first type of pedestrians is greater than the set second threshold, or the overlap time ratio of the target person is greater than the set third threshold, the overlap time period is determined to meet the time overlap standard.

[0151] If the overlap time ratio of the first type of pedestrians is less than or equal to the set second threshold, and the overlap time ratio of the target person is less than or equal to the set third threshold, then the overlap time period is determined not to meet the time overlap standard.

[0152] The second and third thresholds are set independently and can be the same or different.

[0153] For example, with a second threshold of 20% and a third threshold of 30%, if the duration of the overlap between pedestrian K (a type of pedestrian) and target person A in the destination area is t1, the duration of pedestrian K's time in the destination area is t2, and the duration of target person A's time in the destination area is t3, then the overlap ratio of pedestrian K (t1 / t2) and target person A (t1 / t3) is calculated. If t1 / t2 is greater than 20% or t1 / t3 is greater than 30%, the overlap period between pedestrian K and target person A meets the time overlap standard, and pedestrian K is identified as a companion of target person A. If t1 / t2 is less than or equal to 20% and t1 / t3 is less than or equal to 30%, the overlap period between pedestrian K and target person A does not meet the time overlap standard, and pedestrian K is not identified as a companion of target person A.

[0154] It is understandable that, although suspects and their companions are aware of counter-tracking / counter-surveillance, they generally still need to meet in the destination area. Therefore, according to step 140-2, the overlap of stay time in the destination area is analyzed, and the first type of pedestrians identified in step 130 are further screened to select those who meet the time overlap criteria. The behavioral characteristics of these first type of pedestrians are closer to the actual companions of the target person.

[0155] In some exemplary embodiments, step 140-2, which selects the first type of pedestrians whose overlapping time periods meet the time overlap criteria as companions of the target person, further includes:

[0156] If the overlapping time period meets the time overlap standard, the permanent location of the first type of pedestrians corresponding to the overlapping time period is obtained.

[0157] Determine whether the positional relationship between the permanent residence and the destination area meets the proximity criteria. If the proximity criteria are met, determine that the first type of pedestrians corresponding to the overlapping time period are not companions of the target person. If the proximity criteria are not met, determine that the first type of pedestrians corresponding to the overlapping time period are companions of the target person.

[0158] In some exemplary embodiments, the pedestrian's place of residence is derived from pedestrian profile information.

[0159] In some exemplary embodiments, whether the positional relationship satisfies the proximity criterion includes:

[0160] Based on the spatial physical distance between two locations, if the distance is less than the nearest distance threshold, the nearest location standard is determined to be met; if the distance is greater than or equal to the nearest distance threshold, the nearest location standard is determined not to be met.

[0161] Optionally, the determination can also be based on factors such as road connectivity and / or road distance between the two locations. Alternatively, it can be determined whether the endpoint area is within the living activity range of the relevant permanent residents based on the characteristics of their living activity range. If it is, the proximity location criterion is met; otherwise, it is not. Specific determination criteria can be selected or extended to other equivalent variations as needed, based on the above examples, and are not limited to the aspects exemplified in the embodiments of this disclosure.

[0162] For example, if the overlap time between pedestrian K and target A meets the time overlap standard, and based on pedestrian K's permanent residence location, it is determined that pedestrian K's permanent residence location is near the destination area, then although the time overlap between pedestrian K and target A in the destination area is relatively high, the possibility that pedestrian K is a companion of target A is small, and it is uncertain whether they are companions.

[0163] It is understood that in the above embodiments, the residence of the first type of pedestrians who meet the time overlap criteria is investigated, and the first type of pedestrians whose residence is near the destination area are excluded, as they are not considered to be companions. In this way, according to the solution provided by the embodiments of this disclosure, the companions of the target person can be screened out.

[0164] This disclosure also provides a method for identifying companions, used to discover other suspects accompanying a target person (suspect) from pedestrian data collected from multiple dimensions by a monitoring system. These suspects possess counter-tracking and counter-surveillance awareness and behaviors. The method, as described... Figure 4 As shown, it includes:

[0165] Step 410: Establish pedestrian profiles;

[0166] Step 420: Set the target person and their trajectory information;

[0167] Step 430: Obtain the trajectory information of multiple pedestrians that match the trajectory information of the target person;

[0168] Step 440: Trajectory aggregation to determine abnormal trajectory clusters;

[0169] Step 450: Filtering for time overlap in the endpoint region.

[0170] Step 410 includes: obtaining pedestrian profile information from one or more other external systems. Optionally, this could be profile information of the national population, profile information of the local permanent residents, or pedestrian profile information of a specific scope; the specific aspect is not limited.

[0171] Step 430 follows a similar procedure to step 120 above to obtain trajectory information of multiple matching pedestrians. This trajectory information is based on data collected from multiple dimensions of the monitoring system and normalized according to the pedestrians. For example, it includes facial recognition results captured by cameras, visit information of pedestrians carrying terminal devices collected by mobile base stations or Wi-Fi hotspots, or visit information of pedestrians carrying RFID terminals collected by RFID roadside devices, etc., and is not limited to a specific form.

[0172] Step 440 performs trajectory clustering according to the aforementioned step 130, obtaining multiple trajectory clusters, and selecting the clusters that meet the first clustering criteria as abnormal trajectory clusters;

[0173] Step 450: Based on the aforementioned Step 140-2, for the first type of pedestrians who are not the target person in the abnormal trajectory cluster, perform time overlap screening in the endpoint area to determine that the first type of pedestrians who meet the time overlap standard are the companions of the target person.

[0174] If the target's trajectory is included in the abnormal trajectory cluster, it can be understood that the target also has anti-tracking awareness. If the abnormal trajectory cluster determined in step 440 is only one, and the target's trajectory is also included in the abnormal trajectory cluster, it indicates that the target and their companions planned their trajectory in advance and acted according to the planned trajectory route.

[0175] As can be seen, the companion identification scheme provided in this embodiment of the present disclosure, through the collection of multi-dimensional data, concatenates the data by person, determines abnormal trajectories through clustering, and identifies companions with anti-tracking awareness and behavior based on the temporal overlap between the abnormal trajectory and the target object's trajectory, thus solving the problem of companions that cannot be detected by conventional methods. In some exemplary embodiments, during the trajectory object clustering process, spatial similarity, temporal similarity, and trajectory weight deviation are used to screen the first type of pedestrians corresponding to abnormal trajectories, and then further determine the temporal overlap of the endpoint area and the place of residence, eliminating irrelevant data and improving the identification accuracy.

[0176] This disclosure also provides an electronic device, including:

[0177] One or more processors;

[0178] Storage device for storing one or more programs.

[0179] When the one or more programs are executed by the one or more processors, the one or more processors implement the peer identification method as described in any embodiment of this disclosure.

[0180] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, the program being implemented by a processor as the person-fellowship determination method as described in any embodiment of this disclosure.

[0181] The companion identification scheme provided in this disclosure, based on trajectory clustering and endpoint location time overlap determination, effectively identifies companions of a target individual using multi-dimensional pedestrian data collected by the monitoring system. In some embodiments, according to a set first clustering criterion, it is possible to specifically identify suspects and companions with anti-tracking awareness and behavior, solving the problem that it is impossible to effectively detect companions of a target individual under conventional conditions, and further improving the big data analysis function of the security system.

[0182] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0183] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made under the concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A method for identifying fellow travelers, characterized in that, include: Based on the trajectory information of the target person, determine the time range, starting area, and ending area for pedestrian data filtering; Based on the time range, starting area, and ending area, obtain the trajectory information of multiple matching pedestrians; The trajectory objects corresponding to the trajectory information of the multiple pedestrians are aggregated according to the trajectory similarity between the trajectories to obtain multiple trajectory clusters, including: Based on the set road weights, calculate the trajectory weights of each pedestrian's trajectory. The plurality of trajectory objects are used as the initial atomic trajectory objects; Perform the following aggregation and filtering steps on all atomic trajectory objects: The atomic trajectory objects are aggregated according to the trajectory similarity between trajectories to obtain at least one trajectory cluster; the trajectory cluster includes: multi-object trajectory cluster and / or first single-object trajectory cluster; Based on the trajectory weights of the trajectory objects included in each of the multi-object trajectory clusters, trajectory objects whose trajectory weights meet the trajectory offset criteria are removed from their respective multi-object trajectory clusters and treated as a second single-object trajectory cluster. The trajectory objects in the first and second single-object trajectory clusters are used as atomic trajectory objects; Continue performing the aggregation and filtering steps until all trajectory clusters meet the clustering convergence criteria; Select at least one cluster that meets the first clustering criterion from the plurality of trajectory clusters; determine the pedestrians who are not the target person among the pedestrians corresponding to the trajectory objects contained in the selected at least one cluster as the first type of pedestrians; The first type of pedestrians are identified as companions of the target person.

2. The method as described in claim 1, characterized in that, The persons accompanying the first type of pedestrian identified as the target person include: Based on the time period of stay of the first type of pedestrians after arriving at the destination area and the time period of stay of the target person after arriving at the destination area, the overlap time period of each first type of pedestrian and the target person in the destination area is determined, and the first type of pedestrians whose overlap time period meets the time overlap degree standard is selected as the companion of the target person.

3. The method as described in claim 1, characterized in that, in, The trajectory similarity between trajectories is calculated based on the spatial similarity and temporal similarity between trajectories.

4. The method as described in claim 1, characterized in that, The trajectory weights are determined to meet the trajectory offset criteria using the following method: The weight deviation corresponding to the trajectory weight is determined based on the average of the trajectory weights and the clusters to which they belong. If the weight deviation is greater than a preset deviation threshold, the trajectory weight is determined to conform to the trajectory offset standard. If the weight deviation is less than or equal to a preset deviation threshold, it is determined that the trajectory weight does not meet the trajectory offset standard. The mean trajectory weight is the average of the trajectory weights of all pedestrian trajectories in the assigned cluster.

5. The method as described in claim 1, characterized in that, The following method is used to determine whether all trajectory clusters meet the clustering convergence criteria: If the total number of trajectory clusters remains constant, and the trajectory objects contained in each trajectory cluster remain constant, then the clustering convergence criterion is satisfied. If the total number of trajectory clusters changes, or if the trajectory objects contained in any trajectory cluster change, the clustering convergence criterion is determined not to be met.

6. The method according to any one of claims 1-3, characterized in that, Determining whether a trajectory cluster conforms to the first clustering criterion includes: The proportion of trajectory objects included in a trajectory cluster relative to the total number of trajectory objects included in all trajectory clusters is calculated and denoted as the proportion of trajectory objects in a trajectory cluster. If the proportion of trajectory objects is less than a set first threshold, the trajectory clustering is determined to meet the first clustering criterion; if the proportion of trajectory objects is greater than or equal to the set first threshold, the trajectory clustering is determined to not meet the first clustering criterion.

7. The method as described in claim 2, characterized in that, Determining whether the overlapping time period meets the time overlap standard includes: Calculate the proportion of the overlapping time period relative to the dwell time of the corresponding first type of pedestrians after arriving at the destination area, and record it as the overlap time ratio of the first type of pedestrians. Calculate the proportion of the overlapping time period relative to the time the target person stays after arriving at the destination area, and record it as the overlap time ratio of the target person; If the overlap time ratio of the first type of pedestrians is greater than the set second threshold, or the overlap time ratio of the target person is greater than the set third threshold, the overlap time period is determined to meet the time overlap standard. If the overlap time ratio of the first type of pedestrians is less than or equal to the set second threshold, and the overlap time ratio of the target person is less than or equal to the set third threshold, then the overlap time period is determined not to meet the time overlap standard.

8. The method as described in claim 7, characterized in that, The first type of pedestrians selected as those whose overlapping time periods meet the time overlap criteria are companions of the target person, further include: If the overlapping time period meets the time overlap standard, the permanent location of the first type of pedestrians corresponding to the overlapping time period is obtained. Determine whether the positional relationship between the permanent residence and the destination area meets the proximity criteria. If the proximity criteria are met, determine that the first type of pedestrians corresponding to the overlapping time period are not companions of the target person. If the proximity criteria are not met, determine that the first type of pedestrians corresponding to the overlapping time period are companions of the target person.

9. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the peer identification method as described in any one of claims 1-8.

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

Images