Target avoidance behavior pattern recognition method and system

Abstract

The invention discloses a target avoidance behavior pattern recognition method and system, relates to the technical field of target recognition, and aims to solve the problem that an existing target recognition method has obvious limitation under the condition of data missing or deviation caused by man-made interference in position big data. The method comprises the following steps: preprocessing collected suspicious target trajectory data and identifying a stay point; calculating the space-time entropy of the trajectory data, and screening out a target group which may have abnormal behaviors in combination with the characteristics of the spatial data; information black holes in the target group are identified, and target positions and motion paths in the information black holes are mined based on track uncertainty and road network constraint conditions; and in combination with the mined target position and motion path, establishing a recurrence period conceptual model of the space-time event, and carrying out target track space-time anomaly detection. According to the method, comprehensive and missing-free track monitoring can be carried out on the suspicious target, and the dynamic management and control capability of the park is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of target recognition, and in particular to a target avoidance behavior pattern recognition method and system. Background Art

[0002] In some protection parks with higher security requirements, since there is a need to monitor targets entering the park, it is necessary to collect and analyze intelligence on the targets; due to the huge area of ​​the protection park, and some suspicious targets will take hiding measures to deliberately conceal their location and whereabouts, there will be blank areas in the park, resulting in missing or deviation of target trajectory data, and it is impossible to detect trajectory anomalies of suspicious targets, which is not conducive to comprehensive monitoring of the protection park.

[0003] The data sources of existing target recognition methods are directly observable objects, or based on known historical trajectory data, and do not consider the situation where observation records are missing due to hiding measures such as human concealment and disguise, which brings great difficulties to intelligence analysis. For example, the Chinese patent with publication number CN109241223A, which involves a behavior and whereabouts recognition method and platform, does not consider the situation where observation records are missing due to human concealment and disguise. Summary of the invention

[0004] The current target recognition method of the present invention has obvious limitations when human interference in location big data causes data missing or deviation. A target avoidance behavior pattern recognition method and system are proposed. Combining spatiotemporal entropy, motion hypothesis, and recurrence period theory and method, it can conduct comprehensive and non-missing trajectory monitoring of suspicious targets and enhance the dynamic management and control capabilities of the park.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: a method for identifying a target avoidance behavior pattern, comprising the following steps: S1, preprocessing the collected suspicious target trajectory data and identifying the stop points; S2, calculates the spatiotemporal entropy of trajectory data, combines the characteristics of spatial data, and screens out target groups that may have abnormal behaviors; S3, identify the information black holes in the target group, and mine the target position and movement path in the information black holes based on trajectory uncertainty and road network constraints; S4, combining the mined target location and movement path, establishes a conceptual model of the recurrence period of spatiotemporal events to detect spatiotemporal anomalies in the target’s whereabouts.

[0006] In this technical solution, the trajectory data of suspicious targets are first acquired and collected, and then the collected trajectory data are preprocessed and the stop points are identified; then, the given trajectory data is combined to calculate the spatiotemporal entropy of the trajectory data, and the target group that may have abnormal behavior is screened; after that, the information black hole is identified, and the target position and movement path of interest are mined and obtained. In combination with the incomplete trajectory data, a conceptual model of the recurrence period of spatiotemporal time is constructed to explicitly express the hidden data, and the abnormal access location and abnormal access time are identified.

[0007] The present invention is further configured as follows: Step S1 comprises: The trajectory data of suspicious targets are compressed and cleaned, and the spatial areas where the suspicious targets stay longer than the preset threshold are obtained and regarded as stay points, and the key information of each stay point is extracted.

[0008] In this technical solution, after the trajectory data is pre-processed, compressed and cleaned, the stay point is identified, the trajectory of the fixed target is determined by searching, the spatial area that meets the threshold is set as the stay point, and the key information of the stay point is extracted.

[0009] The present invention is further configured as follows: Step S2 comprises: S21, Shannon diversity index was used to calculate spatiotemporal entropy and the SDI value was calculated; S22, determining the spatial range of the trajectory data and calculating the spatial autocorrelation and spatial resolution; S23, construct a quantification method of spatiotemporal entropy, and distinguish target abnormal behavior data set and non-abnormal behavior data through the quantification method of spatiotemporal entropy.

[0010] In this technical solution, since the Shannon diversity index has corresponding limitations in spatial data, the characteristics of spatial data, namely spatial range, spatial autocorrelation and spatial resolution, are combined to construct a quantification method of spatiotemporal entropy. According to this method, the abnormal behavior data set of the target can be identified to determine a clear range for subsequent data analysis.

[0011] The present invention is further configured as follows: Step S3 comprises: S31, uses space-time prism to automatically identify information black holes in target trajectory data; S32, combining the trajectory segment data and the background space elements observable in the information black hole, establishing a motion hypothesis conceptual model, the motion hypothesis conceptual model including the definition and operation of trajectory analysis; S33, combined with the motion hypothesis conceptual model, mines the target location and motion path of interest.

[0012] In this technical solution, after the information black hole is identified, the motion construction concept modeling is carried out through trajectory fragment data and background space elements, and then the target location and motion path of interest are mined; among which, the background space elements include transportation networks and POI points of interest.

[0013] The present invention is further configured as follows: the target whereabouts time-space anomaly detection includes: Routing behavior modeling process,accumulating historical movement data into the trajectory database and building an index to identify normal routing behavior; The anomaly detection process compares real-time information with historical routing behavior knowledge, selects road segments with certain deviations from normal behavior patterns, namely seed segments, and expands each seed segment to a complete anomaly subgraph to determine routing behaviors that have changed significantly.

[0014] In this technical solution, the routing behavior is first modeled, and then the real-time trajectory information is compared with the historical routing behavior knowledge to determine the seed segment and the abnormal subgraph.

[0015] The present invention is further configured such that: the spatial range of the trajectory data is defined by a minimum boundary rectangle, the spatial autocorrelation is calculated by using the Moran index, and the spatial resolution is combined with the spatial range to select a corresponding scale for calculation.

[0016] In the present technical solution, the spatial range can distinguish the size of the effective area, the spatial autocorrelation can indicate the existence of spatial clustered data points, and the spatial resolution can distinguish between fine spatial partitioning and coarse spatial partitioning.

[0017] The present invention is further configured as follows: the information black hole is specifically an area in the trajectory data where observation records are missing due to target hiding or camouflage behavior.

[0018] The present invention is further configured as follows: after step S4, the following steps are further included: Generate anomaly detection information and output it through geospatial intelligence, and determine the interception location and time based on the abnormal access location and abnormal access time.

[0019] In this technical solution, after the abnormal position and time of the target are determined, the abnormal detection information can be output and the interception information can be determined.

[0020] The present invention is further configured as follows: the key information of the stop point includes the average coordinates of the stop point, the arrival time and the stay time.

[0021] In this technical solution, key information of each stop point is determined to complete the stop point identification process.

[0022] A target avoidance behavior pattern recognition system is applicable to the above-mentioned target avoidance behavior pattern recognition method, comprising: The abnormal behavior identification module combines the characteristics of SDI value and spatial data to screen the target group that may have abnormal behavior; Information black hole identification module, which identifies information black holes through space-time prism; Motion hypothesis module, which builds a motion hypothesis conceptual model to mine the target location and motion path of interest; The recurrence period determination module constructs a conceptual model of the recurrence period of spatiotemporal events and conducts spatiotemporal anomaly detection of target whereabouts.

[0023] In the present technical solution, after the trajectory data is preprocessed and the stop points are identified, the data is identified and analyzed in sequence through the abnormal behavior discrimination module, the information black hole identification module, the operation hypothesis module and the recurrence period determination module. The corresponding target group that may have abnormal behavior is obtained according to the abnormal behavior discrimination module, the corresponding regional information is identified through the information black hole identification module, the trajectory concept modeling is carried out through the motion hypothesis module, and the recurrence period concept model is established through the recurrence period hypothesis module, and finally the spatiotemporal anomaly detection of the target's whereabouts is carried out.

[0024] The present invention can bring the following beneficial effects: The present invention relates to a target avoidance behavior pattern recognition method, which combines spatiotemporal entropy, motion hypothesis, and recurrence period theory and method to conduct comprehensive and flawless trajectory monitoring of suspicious targets and enhance the dynamic control capability of the park. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of a method for identifying a target avoidance behavior pattern in the present application.

[0026] Figure 2 It is a schematic diagram of the spatial partition size of a target avoidance behavior pattern recognition method of the present application.

[0027] Figure 3 It is a schematic diagram of an information black hole of a target avoidance behavior pattern recognition method of the present application.

[0028] Figure 4 It is a schematic diagram of a space-time prism of a target avoidance behavior pattern recognition method of the present application.

[0029] Figure 5 It is a schematic diagram of motion path mining of a target avoidance behavior pattern recognition method of the present application.

[0030] Figure 6 It is a schematic diagram of the index structure of the routing behavior modeling process of a target avoidance behavior pattern recognition method of the present application. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] Example 1 This embodiment proposes a method for identifying a target avoidance behavior pattern. Figure 1 , which mainly includes the following steps.

[0033] Step S1, collecting trajectory data of suspicious targets, and preprocessing the collected trajectory data and identifying stop points.

[0034] This step mainly includes the following specific processes: compressing and cleaning the trajectory data; in this embodiment, the trajectory data is GPS trajectory data, which is represented by ASCII text.

[0035] After the trajectory data is compressed and cleaned, the stay points are calculated and obtained. The stay points are the spatial areas where the suspicious target stays for a time greater than a preset threshold. After the stay points are obtained, the key information of the stay points is extracted.

[0036] More specifically, the stop points where the target stays for a period of time in a certain geographical area are calculated. In this embodiment, these stop points can be obtained by searching the GPS track of the target. More specifically, if a target stays within a distance of 200 meters for 30 minutes and the preset threshold is 20 minutes, the area is detected as a stop point.

[0037] The preset threshold can be flexibly adjusted according to actual needs.

[0038] The key information of the stop point mainly includes the average coordinates of the stop point, the arrival time and the stay time; of course, other useful information may also be included, which will not be listed here one by one.

[0039] In this technical solution, after the trajectory data is pre-processed, compressed and cleaned, the stay point is identified, the trajectory of the fixed target is determined by searching, the spatial area that meets the threshold is set as the stay point, and the key information of the stay point is extracted.

[0040] Step S2, after completing the process of step S1, calculate the spatiotemporal entropy of the above trajectory data, and combine the characteristics of spatial data to screen out the target group that may have abnormal behavior. Entropy is a measure of uncertainty, so spatiotemporal entropy can be used to quantify the spatiotemporal predictability of target movement.

[0041] More specifically, step S2 mainly includes the following process.

[0042] Step S21, performing spatiotemporal entropy calculation, specifically using the Shannon diversity index to calculate, and calculating to obtain a corresponding SDI value.

[0043] Step S22, determining the spatial range of the trajectory data, and calculating the spatial autocorrelation and spatial resolution; S23, construct a quantification method of spatiotemporal entropy, and distinguish target abnormal behavior data set and non-abnormal behavior data through the quantification method of spatiotemporal entropy.

[0044] The Shannon Diversity Index is one of the entropy calculation methods, which can essentially quantify the predictability of a target or event.

[0045] For trajectory spatiotemporal data, there is still a certain difference between the SDI value and spatiotemporal predictability. For example, the SDI value is the same when the trajectory data is spatially clustered (spatial predictability is higher) and non-clustered. However, when the underlying data does not change, the difference in the size of the spatial partition (uniform regular grid) will cause the SDI value to change.

[0046] In order to solve the limitations of the Shannon diversity index mentioned above, a quantification method of spatiotemporal entropy is designed based on the spatial attributes, considering the spatial range of spatial data (distinguishing the size of the effective area), spatial autocorrelation (indicating the existence of spatial clustered data points) and spatial resolution (the difference between fine spatial partitioning and coarse spatial partitioning). For example, the size of the spatial partition can refer to Figure 2 .

[0047] More specifically, for the spatial range attribute, the minimum bounding rectangle (MBR) is used for definition; for the spatial autocorrelation attribute, Moran's I index is used for calculation; for the spatial resolution attribute, the corresponding scale is selected for calculation in combination with the size of the spatial range.

[0048] Through the above-mentioned spatiotemporal entropy quantification method, given a set of GPS trajectory spatiotemporal data, the target avoidance behavior is determined by calculating the spatiotemporal entropy and distinguishing the avoidance behavior data set.

[0049] In conclusion, assuming that the target has the same probability of visiting any given location, it is obvious that compared with the daily activities of ordinary people (visiting a small number of fixed locations), the suspect's movement pattern is more complex (usually manifested as visiting more locations in a certain area), and the spatiotemporal entropy value is relatively higher, refer to Table 1: Table 1 Spatiotemporal entropy data of ordinary people and suspects .

[0050] In addition, for the conceptual modeling of space-time entropy, there is the following process: the mathematical measure of information is as follows: Consider the possible output x i , i=1,2,…,n, then event X=x i The self-information of is defined as: , Among them, I(x i ) depends on the base of the logarithm, usually 2. When the base is 2, the unit is bit. Consider a discrete binary source that produces a statistically independent sequence of symbols, with a probability of outputting 0 p and a probability of outputting 1 1-p. The fanning of the binary source is defined as the total amount of information in all possible situations of the binary source: , Generally speaking, the entropy of a fixed N-dimensional random variable X is: , It can be seen that H(X) ≥ 0, when all time probabilities are equal, the entropy of the information source X reaches its maximum, which is logN. The larger the entropy value, the smaller the amount of information it carries, and the smaller the entropy value, the larger the amount of information it carries.

[0051] In addition to the Shannon diversity index, other entropy calculation methods can also be used, including but not limited to Entropy, Tsallis entropy, Kolmogorov-Sinai entropy and Gain entropy.

[0052] Step S3, identifying information black holes in the target group, and mining the target position and movement path in the information black holes based on trajectory uncertainty and road network constraints.

[0053] Step S3 mainly includes the following sub-steps.

[0054] Step S31, using space-time prism to automatically identify information black holes in target trajectory data.

[0055] Information black holes are areas in the trajectory data where observations are missing due to hiding or camouflaging. Assuming that the target is tracked and located persistently, these areas can be detected by identifying significant mismatches between observation expectations and observation records. Expectedly, the trajectory should continuously move through the entire space, with an initial movement trend towards blank areas where observations are missing, but no visit records are observed in these areas. This movement pattern indicates the existence of an information black hole.

[0056] By using the background information of these areas (for example, the known movement patterns of other mobile phone users in the area), the expected continuity of the movement in the area of ​​interest is simulated and estimated, and a conceptual model of the information black hole is established. In addition, in the real world, abnormal situations may occur due to instrument failure or network anomalies. Therefore, it is also necessary to define the type of information black hole (mobile phone blind spots or abnormal areas) and distinguish different abnormal modes (base station outages, some mobile phone users are abnormal). For example, the information black hole can refer to Figure 3 .

[0057] The lack of target observation records can indicate the area of ​​interest. For the area of ​​interest, the space-time prism determines the accessibility of the target's activity location within the observation blank area under time and space constraints, and quickly eliminates irrelevant candidates to automatically identify various information black holes.

[0058] As for the space-time prism, it is an important method of the space-time framework, which geometrically describes the possible range of activities of an individual within a given time interval. Consider a bounded space-time region formed by three orthogonal axes, where the x and y axes represent two-dimensional plane space, and the z axis represents time, located at (x i ,y i ,z i ) is represented by a point target at z i The position of the target in two-dimensional space at any moment. The three-dimensional space range determined by the space-time prism is called the Potential Path Space (PPS), which can measure the expected whereabouts of the observation blank area, with the purpose of estimating the continuity of movement and thus identifying information black holes.

[0059] First, consider that the target's motion is limited to a maximum velocity Vmax between two position point updates. Based on its definition as a geometric set of 2D points, it is clear that the target's position is bounded by an ellipse, with the focus continuously updated at each point position, such as Figure 4 The spatial (XY) projection of is shown in . That is, the assumption is that as long as there are no new (position, time) updates, the target object position can be determined anywhere within the space-time cone based on the last updated position. However, once a new updated position appears, it also constrains the possible positions in the past in addition to the possible future positions.

[0060] Step S32, combining the trajectory segment data and the observable background space elements in the information black hole, to establish a motion hypothesis conceptual model, the motion hypothesis conceptual model includes the definition and operation of trajectory analysis.

[0061] In this technical solution, the observable background spatial elements in the information black hole, such as transportation networks and POI points of interest, are used to make assumptions about the target's whereabouts.

[0062] Step S33, combined with the motion hypothesis conceptual model, the target location and motion path of interest are mined. For the trajectory segment data, the motion hypothesis of multiple sources and multiple destinations (the trajectory segment endpoints around the information black hole) is conceptually expressed and the target location and motion path of interest are mined. Given an existing road network graph and (position, time) samples, all possible trajectory samples that meet the time constraints are constructed. In addition, the probability density function of the position uncertainty along different possible paths is determined.

[0063] More specifically, for the target location, it includes the following: First, since the position of the target is constantly changing, the state of the moving object inside R is divided into sometimes ( ) or always ( ) in [t 1 , t 2 ] in. Secondly, an uncertain moving object, in addition to not being completely inside R, is also 1 , t 2 ] may or definitely satisfy the spatial conditions. In simple terms, if some possible motion curve PMC(Tr) is located in R at time t, it is possible that the object actually moves, or it is possible that the target is already in another PMC(Tr). Formally, the concept of possibly can be specified as , the concept of definitely can also be specified as .

[0064] Quantification is carried out from two aspects: space and time, including: (1) , Semantics: If there is a possible motion curve PMC(Tr) and there is a time t∈[t 1 , t 2 ], so that PMC(Tr) at time t is within region R, then it is true, refer to Figure 5 The first area; (2) , Semantics: For all t∈[t 1 , t 2], if there is a possible motion curve PMC(Tr) such that PMC(Tr) at time t is within region R, then it is true; reference Figure 5 The second area; (3) , Semantics: For every time value t∈[t 1 , t 2 ], true if there exists a possible motion curve PMC(Tr), not necessarily unique, such that PMC(Tr) at time t is within region R or on its boundary; cf. Figure 5 The second area; (4) , Semantics: At every time t∈[t 1 , t 2 ], true if every possible motion curve PMC(Tr) is within region R; (5) , Semantics: For all possible motion curves PMC(Tr), there exists a time t∈[t 1 , t 2 ], where the specific possible motion curve lies in the region R, then it is true; (6) , Semantics: There exists a time point t∈[t 1 , t 2 ], for every possible motion curve PMC(Tr), it is true if it is within the region R.

[0065] For motion path mining, it includes: The same logical formalism is used to analyze the qualitative relationship between range queries and uncertainty trajectories, where the uncertainty model is one of the space-time prism. Not only are the same predicates applicable, but also the relationship between them, Possibly_ Always_Inside Stronger than Always_Possibly_Inside ,as well as Sometime_Definitely_Inside Stronger than Definitely_Sometime_Inside .

[0066] Figure 5 It shows that the two predicates and existential quantifiers are related in the space domain (" Possibly” ) is true for possible motion curves.

[0067] In this technical solution, after the information black hole is identified, the motion construction concept modeling is carried out through trajectory fragment data and background space elements, and then the target location and motion path of interest are mined; among which, the background space elements include transportation networks and POI points of interest.

[0068] Step S4, combining the mined target location and movement path, establish a conceptual model of the recurrence period of spatiotemporal events, and perform spatiotemporal anomaly detection of target whereabouts. Based on the incomplete trajectory data, combined with the target location and movement path of interest obtained through mining, establish a conceptual model of the recurrence period of spatiotemporal events (the interval between revisits) to explicitly express the hidden data.

[0069] For the detection of spatiotemporal anomalies in target whereabouts, it mainly includes the routing behavior modeling process and the anomaly detection process. For the routing behavior modeling process, the process mainly includes accumulating historical movement data into the trajectory database, establishing indexes between road sections and trajectories crossing them, and identifying normal routing behaviors under general circumstances.

[0070] For the anomaly detection process, first, the received GPS trajectories are mapped to the road network using a map matching algorithm. These processed trajectories are used for real-time routing behavior analysis. By comparing real-time information with historical routing behavior knowledge, road segments with certain deviations from normal behavior patterns are selected as seed segments. Each seed segment is expanded to a complete anomaly subgraph to determine the routing behavior that has changed significantly.

[0071] The routing behavior modeling process is more detailed and includes the following processes: To detect anomalies, it is necessary to efficiently find all the set segments on the trajectory and check their routing behavior. Therefore, two index structures are created to speed up the anomaly detection process. Figure 6 As shown, the directed solid lines represent road segments and the dotted lines represent trajectories. The left side is a bidirectional index structure between trajectories and road segments. Each trajectory is converted into a path of connected road segments. The forward index is between each different trajectory and all the road segments contained in the derived path. Each road segment is indexed and traversed according to each trajectory. The right index is to establish an index structure for each road segment r to index all the end paths and trajectories along them. The end path refers to the path containing r as the last edge. However, if the trajectories of all possible road segment combinations in the road network are indexed, the size of the index entries grows exponentially relative to the size of the road network. Therefore, in the anomaly detection process, the left index is used to build an additional index structure to speed up the search process.

[0072] The anomaly detection process includes the following steps: Given a set of road segments R = {r 1 ,r 2 ,…, r n}、A set of drivers U={u 1 ,u 2 ,…,u m} and a group in [t 0 ,t1 ] between the trajectory TR = {tr 1 ,tr 2 ,…,tr h}, find a set of subgraphs, called abnormal graphs {g 1 ,g 2 ,…,g k}, where each graph g i Meet the following criteria: (a) g i is a connected graph; (b) In g i For each r O , at least one r D , in r O and r D T 1 The routing pattern (RP) at the location meets the following conditions: , in, Indicates that in [t 0 ,t 1 ) is the median of all routing patterns calculated in . Similarly, for each r D , at least one r O The above constraints are satisfied. The left side of the formula is t 1 The Mahalanobis distance between the routing pattern and the regular routing pattern. On the right is t 1 Three times the standard deviation of all previous routing patterns. Therefore, the goal is to detect anomalous routing patterns that highly deviate from regular routing patterns. To do this, a seed segment is selected and expanded into a subgraph that satisfies the above conditions.

[0073] Anomaly seed segment: At time T, an anomaly seed segment is a segment that does not exist in the existing anomaly subgraph and has a flow f that satisfies the following equation T : , μ is the historical median flow rate in the road segment, t 1 to N Refers to all timestamps before T. At time T, all road segments that meet the above definition are put into the abnormal seed candidate pool. The road segments in the pool are sorted and arranged according to the size of the traffic change. To ensure that the selected seeds are not in any existing anomaly graph, select a seed at a time and expand the corresponding anomaly graph, and then repeat this process.

[0074] After step S4, the following steps are also included: generating abnormal detection information and outputting it through geospatial intelligence, and determining the interception location and time based on the abnormal access location and abnormal access time.

[0075] This embodiment can bring the following technical effects.

[0076] 1. Aiming at the spatiotemporal trajectory of the target in the location big data, combined with the Shannon diversity index in information theory, by analyzing the limitations of the index in spatial data, a spatiotemporal entropy conceptual model is established to express the spatiotemporal predictability of the target movement. Combined with the spatiotemporal entropy conceptual model, the spatial range, spatial autocorrelation, and spatial resolution characteristics of spatial data are applied to establish a quantitative method for spatiotemporal entropy. By calculating the spatiotemporal entropy, the target abnormal and non-abnormal behavior data sets are distinguished to determine the scope for subsequent data analysis.

[0077] 2. For the distinguished target abnormal behavior data set, based on the expected continuity characteristics of the movement, the information black hole conceptual model is established to express the areas of interest where the observation records are missing in the target abnormal behavior, and then the types of information black holes are defined to distinguish different abnormal behavior patterns. In combination with the information black hole conceptual model, the information black hole mining algorithm is set up to analyze the areas of interest where the observation records are missing by calculating the accessibility of the activity places of individual behavior under time and space constraints, thereby identifying information black holes.

[0078] 3. Based on the determination of the information black hole, the trajectory fragment data around the information black hole and the observable background space elements (traffic network, POI points of interest) in the information black hole are used to establish a motion hypothesis conceptual model to express the possible position and motion path of the target in the information black hole.

[0079] 4. Based on the movement hypothesis, a conceptual model of the return period of target spatiotemporal events is established for possible paths, main channels and key locations where the target intends to hide, to express the interval frequency of the target's visit to the area of ​​interest. Combined with the conceptual model of the return period, a method for detecting spatiotemporal anomalies in target whereabouts is established to quantify and identify spatiotemporal anomalies in the target whereabouts dataset (abnormal access locations and abnormal access times corresponding to daily activities), thereby predicting the target's schedule to help interception and surveillance.

[0080] Example 2 Based on Example 1, this example further proposes a target avoidance behavior pattern recognition system, including an abnormal behavior discrimination module, an information black hole recognition module, a motion hypothesis module, and a recurrence period determination module.

[0081] Among them, the abnormal behavior identification module combines the characteristics of SDI value and spatial data, and through this module, it is possible to screen out the target group that may have abnormal behavior.

[0082] The information black hole identification module can identify information black holes through the space-time prism, and the motion hypothesis module can establish a motion hypothesis conceptual model and ultimately discover the target location and running path of interest.

[0083] The return period determination module can establish a conceptual model of the return period based on space-time and perform space-time anomaly detection of target whereabouts.

[0084] In the present technical solution, after the trajectory data is preprocessed and the stop points are identified, the data is identified and analyzed in sequence through the abnormal behavior discrimination module, the information black hole identification module, the operation hypothesis module and the recurrence period determination module. The corresponding target group that may have abnormal behavior is obtained according to the abnormal behavior discrimination module, the corresponding regional information is identified through the information black hole identification module, the trajectory concept modeling is carried out through the motion hypothesis module, and the recurrence period concept model is established through the recurrence period hypothesis module, and finally the spatiotemporal anomaly detection of the target's whereabouts is carried out.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A method for identifying a target avoidance behavior pattern, characterized in that: The following steps are involved: S1, preprocessing the collected suspicious target trajectory data and identifying the stop points; S2, calculates the spatiotemporal entropy of trajectory data, combines the characteristics of spatial data, and screens out target groups that may have abnormal behaviors; S3, identify the information black holes in the target group, and mine the target position and movement path in the information black holes based on trajectory uncertainty and road network constraints; S4, combining the mined target location and movement path, establishes a conceptual model of the recurrence period of spatiotemporal events to detect spatiotemporal anomalies in the target’s whereabouts.

2. A method for identifying a target avoidance behavior pattern according to claim 1, characterized in that: The step S1 comprises: The trajectory data of suspicious targets are compressed and cleaned, and the spatial areas where the suspicious targets stay longer than the preset threshold are obtained and regarded as stay points, and the key information of each stay point is extracted.

3. A method for identifying a target avoidance behavior pattern according to claim 1 or 2, characterized in that: The step S2 comprises: S21, Shannon diversity index was used to calculate spatiotemporal entropy and the SDI value was calculated; S22, determining the spatial range of the trajectory data and calculating the spatial autocorrelation and spatial resolution; S23, construct a quantification method of spatiotemporal entropy, and distinguish target abnormal behavior data set and non-abnormal behavior data through the quantification method of spatiotemporal entropy.

4. A method for identifying a target avoidance behavior pattern according to claim 3, characterized in that: The step S3 comprises: S31, uses space-time prism to automatically identify information black holes in target trajectory data; S32, combining the trajectory segment data and the background space elements observable in the information black hole, establishing a motion hypothesis conceptual model, the motion hypothesis conceptual model including the definition and operation of trajectory analysis; S33, combined with the motion hypothesis conceptual model, mines the target location and motion path of interest.

5. The target avoidance behavior pattern recognition method according to claim 1, characterized in that: The target whereabouts spatiotemporal anomaly detection includes: Routing behavior modeling process,accumulating historical movement data into trajectory database and building index to identify normal routing behavior; The anomaly detection process compares real-time information with historical routing behavior knowledge, selects road segments with certain deviations from normal behavior patterns, namely seed segments, and expands each seed segment to a complete anomaly subgraph to determine routing behaviors that have changed significantly.

6. A method for identifying a target avoidance behavior pattern according to claim 3, characterized in that: The spatial range of the trajectory data is defined by a minimum bounding rectangle, the spatial autocorrelation is calculated by using the Moran index, and the spatial resolution is combined with the spatial range to select a corresponding scale for calculation.

7. A method for identifying target avoidance behavior patterns according to claim 4, characterized in that: The information black hole is specifically an area in the trajectory data where observation records are missing due to target hiding or camouflage behavior.

8. A method for identifying a target avoidance behavior pattern according to claim 1, 2 or 4, characterized in that: After step S4, the following steps are also included: Generate anomaly detection information and output it through geospatial intelligence, and determine the interception location and time based on the abnormal access location and abnormal access time.

9. A method for identifying target avoidance behavior patterns according to claim 2, characterized in that: The key information of the stay point includes the average coordinates of the stay point, the arrival time and the stay time.

10. A target avoidance behavior pattern recognition system, applicable to a target avoidance behavior pattern recognition method according to any one of claims 1 to 9, characterized in that: include: The abnormal behavior identification module combines the characteristics of SDI value and spatial data to screen the target group that may have abnormal behavior; Information black hole identification module, which identifies information black holes through space-time prism; Motion hypothesis module, which builds a motion hypothesis conceptual model to mine the target location and motion path of interest; The recurrence period determination module constructs a conceptual model of the recurrence period of spatiotemporal events and conducts spatiotemporal anomaly detection of target whereabouts.

Images