A method and system for matching multi-source sensing data based on spatial correlation
By constructing spatial association graphs and describing local spatial structures, the problem of unstable target object matching in multi-source perception systems is solved, achieving efficient and stable matching in complex scenarios and reducing the false matching rate.
Patent Information
- Application Number
- CN202611292164.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-10
AI Technical Summary
In existing multi-source sensing systems, it is difficult to stably match target objects between different heterogeneous sensing sources, which can easily lead to target mismatch, target loss, or unstable association, especially in complex dynamic scenarios where the matching error rate is high.
A spatial association graph is constructed, and heterogeneous sensing data is mapped through a unified spatial reference system to generate a target spatial distribution set. Based on the spatial association graph, a local spatial association domain is generated. Through association candidate propagation processing and spatial structure consistency detection, conflicting target objects are corrected to achieve stable matching of multi-source sensing data.
It improves matching stability in complex scenarios, reduces false matching rate, and increases processing efficiency. By constructing spatial association graphs and local spatial structure descriptions, it achieves structured matching of multi-source sensing targets.
Smart Images

Figure CN123068795A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of computer technology, specifically relating to a multi-source sensing data matching method and system based on spatial correlation. Background Technology
[0002] With the development of technologies such as smart cities, intelligent warehousing, autonomous driving, intelligent transportation, and the Industrial Internet of Things, multi-source heterogeneous sensing systems are widely used for target detection and state perception in complex scenarios. Existing systems typically deploy multiple heterogeneous sensing devices simultaneously, such as video sensing devices, positioning sensing devices, radio frequency identification devices, and environmental sensing devices, to achieve multi-dimensional perception of target objects.
[0003] However, due to differences in sensing methods, sampling frequencies, spatial precision, and data representation formats, it is difficult to stably match target objects in different sensing data from different heterogeneous sensing sources. For example, video sensing devices acquire image location information of target objects, RFID sensing devices acquire tag identification information, and positioning sensing devices acquire coordinate trajectory information. There is usually no unified target identifier among different sensing sources, which can easily lead to problems such as target mismatch, target loss, or unstable association.
[0004] Most existing multi-source data matching methods employ time synchronization, coordinate distance matching, or single-feature similarity matching for target association. For example, they determine whether targets belong to the same object based on Euclidean distance, or align data from multiple sensor sources based on timestamps. While these methods can achieve basic target matching in simple scenarios, they are prone to target association errors in complex and dynamic scenarios.
[0005] Application content
[0006] The purpose of this application is to provide a multi-source sensing data matching method and system based on spatial correlation, so as to solve the defect of the prior art that is prone to target correlation errors.
[0007] To solve the above-mentioned technical problems, this application is implemented as follows:
[0008] Firstly, a multi-source sensing data matching method based on spatial correlation is provided, including the following steps:
[0009] The system acquires sensing data from multiple heterogeneous sensing sources, and performs spatial mapping processing on the target objects detected by the sensing data from each heterogeneous sensing source based on a unified spatial reference system to form a target spatial distribution set; the multiple heterogeneous sensing devices include multiple devices from video sensing devices, positioning sensing devices, radio frequency identification devices, and environmental sensing devices.
[0010] A spatial association graph is constructed based on the target spatial distribution set. The nodes in the spatial association graph represent the target objects detected by each sensing data, and the edges in the spatial association graph represent the spatial association relationships between different target objects. The spatial association relationships include adjacency relationships, directional association relationships, motion continuity relationships, and spatial dependency relationships.
[0011] For the target object to be matched corresponding to the first heterogeneous sensing source among the plurality of heterogeneous sensing sources, a local spatial association domain is generated based on the spatial association graph. The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target direction distribution state, and the target spatial attachment state.
[0012] Based on the local spatial association domain, the association candidate propagation process is performed to generate a propagation candidate set. The association candidate propagation process is used to extend and propagate the candidate target objects corresponding to the heterogeneous sensing sources other than the first heterogeneous sensing source along the spatial association graph according to the spatial association relationship. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path.
[0013] A spatial structure consistency detection is performed on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is a spatial structure conflict between target objects corresponding to different heterogeneous sensing sources. The spatial structure conflict includes target intersection conflict, movement direction conflict, space occupation conflict and spatial dependency contradiction conflict.
[0014] When a spatial structure conflict is detected, the conflicting target object is subjected to association correction processing based on the continuity of the propagation path, and a matching result corresponding to the target object to be matched is generated based on the propagation candidate set after association correction.
[0015] Secondly, a multi-source sensing data matching system based on spatial correlation is provided, including:
[0016] The acquisition module is used to acquire sensing data from multiple heterogeneous sensing sources, and perform spatial mapping processing on the target objects detected by the sensing data of each heterogeneous sensing source based on a unified spatial reference system to form a target spatial distribution set; the multiple heterogeneous sensing devices include multiple devices from video sensing devices, positioning sensing devices, radio frequency identification devices and environmental sensing devices;
[0017] The construction module is used to construct a spatial association graph based on the target spatial distribution set. The nodes in the spatial association graph represent target objects detected by each sensing data, and the edges in the spatial association graph represent spatial association relationships between different target objects. The spatial association relationships include adjacency relationships, directional association relationships, motion continuity relationships, and spatial dependency relationships.
[0018] The first generation module is used to generate a local spatial association domain based on the spatial association graph for the target object to be matched corresponding to the first heterogeneous sensing source among the plurality of heterogeneous sensing sources. The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target direction distribution state, and the target spatial attachment state.
[0019] The second generation module is used to perform association candidate propagation processing based on the local spatial association domain to generate a propagation candidate set. The association candidate propagation processing is used to extend and propagate the candidate target objects corresponding to the heterogeneous sensing sources other than the first heterogeneous sensing source along the spatial association graph according to the spatial association relationship. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path.
[0020] The detection module is used to perform spatial structure consistency detection on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is spatial structure conflict between target objects corresponding to different heterogeneous sensing sources. The spatial structure conflict includes target intersection conflict, movement direction conflict, space occupancy conflict and spatial dependency contradiction conflict.
[0021] The correction module is used to perform association correction processing on the conflicting target object according to the continuity of the propagation path when a spatial structure conflict is detected, and to generate a matching result corresponding to the target object to be matched based on the propagation candidate set after association correction.
[0022] This application's embodiments construct a spatial association graph and introduce dynamic propagation weights of distance, direction, and spatial attachment to achieve structured matching of multi-source sensing targets. By combining local spatial structure description and propagation consistency detection, the matching stability in complex scenarios can be improved. Through conflict correction and hierarchical compression, the false matching rate can be effectively reduced and the processing efficiency can be improved. Attached Figure Description
[0023] Figure 1 This is a flowchart of a multi-source sensing data matching method based on spatial correlation provided in an embodiment of this application;
[0024] Figure 2 This is a schematic diagram of the spatial distribution of multiple target objects in a unified spatial coordinate system provided in the embodiments of this application;
[0025] Figure 3 This is a local association subgraph provided in an embodiment of this application;
[0026] Figure 4 This is another local association subgraph provided in the embodiments of this application;
[0027] Figure 5 This is a schematic diagram of the structure of a multi-source sensing data matching system based on spatial correlation provided in an embodiment of this application. Detailed Implementation
[0028] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0029] like Figure 1 The diagram shown is a flowchart of a multi-source sensing data matching method based on spatial correlation provided in an embodiment of this application. The method includes the following steps:
[0030] Step 101: Acquire sensing data from multiple heterogeneous sensing sources, and perform spatial mapping processing on the target objects detected by the sensing data of each heterogeneous sensing source based on a unified spatial reference frame to form a target spatial distribution set.
[0031] The plurality of heterogeneous sensing devices include multiple devices from video sensing devices, positioning sensing devices, radio frequency identification devices, and environmental sensing devices.
[0032] For example, in a smart warehousing scenario, video sensing devices, RFID sensing devices, UWB positioning devices, and environmental sensing devices are deployed within the warehousing area. Video sensing devices detect personnel targets within the warehousing area, RFID sensing devices identify cargo tags, UWB positioning devices detect handling vehicles, and environmental sensing devices detect temperature and humidity changes and occupancy status in localized areas. The system first acquires sensing data from various heterogeneous sensing sources and performs spatial mapping processing on the target objects detected by each heterogeneous sensing source based on a unified spatial reference frame, thereby forming a set of target spatial distributions. The target object data detected by each heterogeneous sensing device at multiple consecutive time points are as follows:
[0033] At time t1, the video sensing device detects personnel A and E, with coordinates (2.0, 3.0) and (8.5, 2.0) respectively; the RFID sensing device detects goods B and D, with coordinates (3.1, 4.0) and (6.6, 5.4) respectively; the UWB positioning device detects the transport vehicle C, with coordinates (6.2, 5.0); and the environmental sensing device detects the area temperature control node F and the area occupancy node G, with coordinates (3.5, 3.8) and (6.8, 5.6) respectively. The spatial distribution of each of these target objects corresponds to a unified spatial coordinate system, as follows: Figure 2 As shown. At time t2, the coordinates of personnel A and E are (2.5, 3.2) and (5.0, 7.0), respectively; the coordinates of goods B and D are (3.5, 4.2) and (7.2, 5.7), respectively; the coordinates of transport vehicle C are (6.8, 5.3); and the coordinates of area temperature control node F and area occupancy node G are (3.6, 3.9) and (7.0, 5.9), respectively. At time t3, the coordinates of personnel A and E are (3.0, 3.5) and (9.0, 1.5), respectively; the coordinates of goods B and D are (4.0, 4.5) and (7.9, 6.2), respectively; the coordinates of transport vehicle C are (7.5, 5.8); and the coordinates of area temperature control node F and area occupancy node G are (3.8, 4.1) and (7.5, 6.3), respectively.
[0034] Step 102: Construct a spatial association graph based on the target spatial distribution set. The nodes in the spatial association graph represent the target objects detected by each sensing data, and the edges in the spatial association graph represent the spatial association relationships between different target objects.
[0035] Spatial relationships include adjacency, directional relationships, motion continuity, and spatial dependence.
[0036] Specifically, spatial neighborhoods can be divided for each target object in the target spatial distribution set to generate multiple local spatial sub-regions; based on the spatial positional relationships, movement direction relationships, and spatial dependency relationships between multiple target objects in each local spatial sub-region, a corresponding local association subgraph is constructed; each node in the local association subgraph represents each target object within the local spatial sub-region, and the edges in the local association subgraph represent the spatial association relationships between different target objects; according to the spatial change trends of two target objects connected by the same edge in the local association subgraph over multiple consecutive time periods, a corresponding dynamic propagation weight is assigned to the edge connecting the two target objects.
[0037] In this embodiment, the spatial distance correlation parameter, motion direction correlation parameter, and spatial attachment correlation parameter between the two target objects can be calculated using the following formulas:
[0038]
[0039] Among them, R ij is the spatial distance correlation parameter between target object i and target object j, used to characterize their spatial proximity and stability in consecutive time intervals; T is the time window length, representing the total number of consecutive time intervals used for calculation; t is the time index, representing the t-th time interval within the time window length; d ij (t) represents the spatial distance between target object i and target object j at time t; Δd ij (t) represents the change in spatial distance between target object i and target object j between adjacent time points; σ is the distance decay coefficient, used to control the decay rate of the influence of spatial distance on the association strength, and the distance decay coefficient is determined based on the average spacing between target objects; η is the distance change stability coefficient, used to control the degree of influence of distance fluctuation on the association strength, and the distance change stability coefficient is determined based on the variance of distance change between target objects;
[0040]
[0041] Among them, V ij θ is the motion direction association parameter between target object i and target object j; T is the time window length, representing the total number of consecutive moments used for calculation; t is the time index, representing the t-th moment in the time window length; ij (t) represents the angle between the motion directions of target object i and target object j at time t; Δθ ij (t) represents the change in the angle between the motion directions of target object i and target object j between adjacent time points; κ is the direction stability decay coefficient, which is used to control the influence of direction changes on the association strength. The direction stability decay coefficient is determined based on the degree of dispersion of the motion directions between target objects.
[0042]
[0043] Among them, A ij is the spatial dependency parameter between target object i and target object j, used to characterize whether a stable spatial binding relationship exists between them; T is the time window length, representing the total number of consecutive moments used for calculation; t is the time index, representing the t-th moment in the time window length; d ij (t) represents the spatial distance between target object i and target object j at time t; Δr ij(t) represents the change in the relative position between target object i and target object j between adjacent time points, used to characterize spatial attachment stability; σ is the distance decay coefficient, used to control the decay rate of the influence of spatial distance on the association strength; μ is the relative position stability coefficient, used to control the influence of spatial structure stability, which is determined according to the magnitude of the relative position change between target objects.
[0044] It should be noted that the distance attenuation coefficient, distance change stability coefficient, direction stability attenuation coefficient, and relative position stability coefficient are all determined based on the spatial relationship sequence formed by the target object pair consisting of the first target object and the second target object at consecutive times. This spatial relationship sequence includes the distance sequence, direction sequence, and relative position sequence.
[0045] Specifically, the distance attenuation coefficient is determined as follows: a first target object and a second target object are used to form a target object pair. The spatial positions of the two objects at each time point are obtained within a preset time window, and the spatial distance sequence of the target object pair at each time point is calculated. The spatial distance sequence is statistically analyzed to obtain statistical values that characterize the spatial proximity scale characteristics of the target object pair, and the distance attenuation coefficient is determined based on the statistical values.
[0046] The distance change stability coefficient is determined as follows: a first target object and a second target object are used to form a target object pair. The spatial distance change of the target object pair is calculated at continuous time intervals, and the distance change is accumulated and statistically analyzed within a preset time window to obtain a parameter characterizing the degree of fluctuation of the spatial distance change of the target object pair. The parameter is then used as the distance change stability coefficient.
[0047] The method for determining the directional stability attenuation coefficient is as follows: a first target object and a second target object are used to form a target object pair. The motion trajectories of the two objects at consecutive moments are obtained, and the relative motion direction angle sequence of the target object pair is calculated based on the motion trajectory. The changing trend of the angle sequence is statistically analyzed to obtain a parameter characterizing the consistency and stability of the motion direction of the target object pair, and the parameter is used as the directional stability attenuation coefficient.
[0048] The relative position stability coefficient is determined as follows: a target object pair is formed by the first target object and the second target object. The relative position vector between the two is calculated at consecutive time intervals. The change of the relative position vector in the time dimension is statistically processed to obtain a parameter characterizing the degree of stability of the target object in the spatial structure. The parameter is used as the relative position stability coefficient.
[0049] Furthermore, based on the spatial distance association parameters, motion direction association parameters, and spatial attachment association parameters between the two target objects, corresponding dynamic propagation weights are assigned to the edges connecting the two target objects. These dynamic propagation weights characterize the stability of the spatial association between different target objects and are calculated using the following formula:
[0050]
[0051] Among them, W ij R represents the dynamic propagation weights between target object i and target object j. ij V is the spatial distance correlation parameter between target object i and target object j. ij A is the motion direction association parameter between target object i and target object j. ij Let λ1, λ2, and λ3 be the spatial dependency parameters between target object i and target object j. Let λ1, λ2, and λ3 be the weight coefficients corresponding to the spatial distance parameter, motion direction parameter, and spatial dependency parameter, respectively. The values of λ1, λ2, and λ3 are all greater than zero and less than 1, and λ1+λ2+λ3=1.
[0052] In this embodiment, the values of the weighting coefficients λ1, λ2 and λ3 can be determined based on the spatial relationship sequence formed by the target object pair within a preset time window.
[0053] Specifically, a target object pair is formed by the first target object and the second target object. The spatial distance correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the spatial distance correlation parameter sequence is statistically analyzed to obtain a distance correlation stability index characterizing the stability of spatial distance correlation. The distance correlation stability index is then normalized to obtain a weighting coefficient λ1. Similarly, a target object pair is formed by the first target object and the second target object. The motion direction correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the motion direction correlation parameter sequence is statistically analyzed to obtain a direction correlation stability index characterizing the stability of motion direction correlation. The direction correlation stability index is then normalized to obtain a weighting coefficient λ2. Finally, a target object pair is formed by the first target object and the second target object. The spatial dependency correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the spatial dependency correlation parameter sequence is statistically analyzed to obtain a spatial dependency stability index characterizing the stability of spatial dependency correlation. The spatial dependency stability index is then normalized to obtain a weighting coefficient λ3.
[0054] Subsequently, based on the boundary overlap relationship between adjacent local spatial sub-regions, each local association subgraph is connected to generate a global spatial association graph. When there is a common target object or continuous spatial propagation path between adjacent local spatial sub-regions, cross-regional connection edges are established between the corresponding local association subgraphs. Based on the local spatial aggregation state of each target object, hierarchical association compression processing is performed on the global spatial association graph to generate a spatial association graph containing core association nodes and propagation association edges. The hierarchical association compression processing is used to associate and aggregate multiple target objects that are spatially continuous and have the same direction of movement, and to prune connection edges whose association stability is lower than a preset threshold.
[0055] In this embodiment, the system constructs a spatial association graph based on the target spatial distribution set formed in step 101. First, the system divides the spatial neighborhood of each target object in the target spatial distribution set and generates multiple local spatial sub-regions. For example, for... Figure 2 The spatial distribution shown can generate local spatial sub-regions R1: {A, B, F}, R2: {C, D, G}, and R3: {E}. Local spatial sub-region R1 corresponds to the racking operation area, local spatial sub-region R2 corresponds to the loading and unloading transportation area, and local spatial sub-region R3 corresponds to the inspection activity area.
[0056] Subsequently, the system constructs corresponding local association subgraphs based on the spatial positional relationships, movement direction relationships, and spatial dependency relationships between target objects in each local spatial sub-region. For example... Figure 3 The diagram shows the local association subgraph corresponding to the local spatial sub-region R1. A and B have a stable transport relationship, B and F have a fixed shelving area relationship, and A and F have a spatial proximity relationship. Figure 4 The diagram shows the local relational subgraph corresponding to the local spatial sub-region R2. There is a transport dependency relationship between C and D, a region occupancy relationship between D and G, and a spatial synchronous movement relationship between C and G.
[0057] Subsequently, based on the spatial change trends of two target objects connected by the same edge over multiple consecutive time points, the system calculates the spatial distance correlation parameter, motion direction correlation parameter, and spatial dependency correlation parameter. For example, for target object A and target object B, the system calculates: spatial distance correlation parameter 0.86, motion direction correlation parameter 0.97, spatial dependency correlation parameter 0.91, and dynamic propagation weight 0.91. For target object C and target object D, the system calculates: spatial distance correlation parameter 0.89, motion direction correlation parameter 0.95, spatial dependency correlation parameter 0.96, and dynamic propagation weight 0.94. For target object A and target object E, the spatial distance correlation parameter is 0.21, motion direction correlation parameter is 0.13, spatial dependency correlation parameter is 0.09, and dynamic propagation weight is 0.14.
[0058] The system connects the local association subgraphs according to the boundary overlap relationships between adjacent local spatial sub-regions and generates a global spatial association graph. Subsequently, based on the local spatial clustering state of the target object, the system performs hierarchical association compression processing on the global spatial association graph. For example, aggregation cluster 1: {A, B, F}; aggregation cluster 2: {C, D, G}. Here, aggregation cluster 1 represents the shelf handling association structure, and aggregation cluster 2 represents the transportation loading and unloading association structure. Because target E has low association stability, edges AE and BE are clipped.
[0059] Step 103: For the target object to be matched corresponding to the first heterogeneous sensing source among the multiple heterogeneous sensing sources, a local spatial association domain is generated based on the spatial association graph.
[0060] The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target orientation distribution state, and the target spatial dependency state. Accordingly, using the target object to be matched as the central node, association nodes with a dynamic propagation weight higher than a preset association threshold are searched in the spatial association graph, and a local spatial association domain is formed based on these association nodes. According to the target object distribution state in the local spatial association domain, a target density distribution model, a target orientation distribution model, and a target spatial dependency distribution model are constructed. The target density distribution model characterizes the degree of aggregation of target objects within the local spatial range, the target orientation distribution model characterizes the consistency of movement directions between target objects, and the target spatial dependency distribution model characterizes the spatial binding stability between target objects. Based on the target density distribution model, the target orientation distribution model, and the target spatial dependency distribution model, a local spatial structure description vector corresponding to the target object to be matched is generated to characterize the spatial structure state around the target object.
[0061] For example, the system uses target object A as the target object to be matched and sets an association threshold of 0.7. The system searches the spatial association graph for association nodes with a dynamic propagation weight higher than 0.70 between them and target A, ultimately forming a local spatial association domain: Local(A) = {A, B, F}. Subsequently, the system constructs a target density distribution model. Since the number of target objects in Local(A) is large and their distribution is compact, the target density is high. The system further analyzes the movement directions in consecutive time steps: A to the right, B to the right, and F to the right. Since multiple target objects move in the same direction, the direction consistency is high. Simultaneously, A and B maintain a stable distance over a long period, and B and F remain in a fixed shelf area for a long time; therefore, the spatial attachment stability is high. Subsequently, the system generates a local spatial structure description vector corresponding to target object A:
[0062] V_local(A)=
[0063] {
[0064] Target density: High
[0065] Directional consistency: High
[0066] Spatial attachment stability: High
[0067] }
[0068] Step 104: Perform association candidate propagation processing based on the local spatial association domain to generate a propagation candidate set.
[0069] The associated candidate propagation process is used to propagate the candidate target objects corresponding to the other heterogeneous sensing sources (excluding the first heterogeneous sensing source) along the spatial association graph based on spatial association relationships. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path.
[0070] Specifically, the target object to be matched is used as the initial propagation node, and multi-level extended propagation is performed along the propagation association edges in the spatial association graph. During the propagation process, the dynamic propagation weights corresponding to each propagation association edge in the propagation path are accumulated to obtain the path propagation credibility. When the path propagation credibility is higher than the preset propagation credibility threshold, the corresponding candidate target object is retained as a propagation node to continue to perform extended propagation, and a propagation candidate set is generated according to the propagation path length, propagation level depth and path propagation credibility of each candidate target object.
[0071] For example, the system sets the propagation trust threshold to 0.75 and uses target object A as the initial propagation node, performing multi-level extended propagation along the propagation association edges: A → B, A → F. Since the dynamic propagation weight between target object B and target object C is 0.26, and the dynamic propagation weight between target object F and target object C is 0.19, both are below the propagation trust threshold, propagation stops. The system then accumulates the dynamic propagation weights along the propagation path.
[0072] Path(A→B)=0.91
[0073] Path(A→F)=0.82
[0074] Therefore, the propagation candidate set is: {B, F}.
[0075] Step 105: Perform spatial structure consistency detection on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is a spatial structure conflict between target objects corresponding to different heterogeneous sensing sources.
[0076] Spatial structural conflicts include target intersection conflicts, movement direction conflicts, space occupancy conflicts, and spatial dependency conflicts. Accordingly, each candidate target object in the propagation candidate set can be obtained. Using each candidate target object as a central node, the spatial association graph searches for association nodes with dynamic propagation weights higher than a preset association threshold to generate local spatial association domains for the corresponding candidate target objects. Based on the target object distribution state in the local spatial association domains corresponding to each candidate target object, a target density distribution model, a target direction distribution model, and a target spatial dependency distribution model are constructed. Based on these models, local spatial structure description vectors for the corresponding candidate target objects are generated. Based on these local spatial structure description vectors, a spatial structure consistency constraint model is constructed to detect whether spatial structural conflicts exist between different candidate target objects. These spatial structural conflicts include: spatial position intersection conflicts, movement direction reversal conflicts, overlapping space occupancy area conflicts, and spatial dependency relationship breakage conflicts between target objects. Based on the spatial structure change trends of each candidate target object in continuous time intervals, it is determined whether spatial structural conflicts exist between candidate target objects.
[0077] In this embodiment, a spatial structure consistency score can be calculated based on the spatial structure change trend of each candidate target object in continuous time. When the spatial structure consistency score is lower than a preset consistency threshold, it is determined that there is a spatial structure conflict between the corresponding candidate target objects.
[0078] For example, the system generates corresponding local spatial association domains with target object B and target object F as central nodes: Local(B)={A,B,F}, Local(F)={A,B,F}, and generates local spatial structure description vectors for target object B, target object F, and target object E respectively.
[0079] V_local(B)=
[0080] {
[0081] Target density: High
[0082] Directional consistency: High
[0083] Spatial attachment stability: High
[0084] }
[0085] V_local(F) =
[0086] {
[0087] Target density: High
[0088] Directional consistency: High
[0089] Spatial attachment stability: High
[0090] }
[0091] V_local(E)=
[0092] {
[0093] Target density: Low
[0094] Directional consistency: Low
[0095] Spatial attachment stability: low
[0096] }
[0097] Subsequently, the system constructs a spatial structure consistency constraint model for V_local(A) and V_local(B).
[0098] By comparing V_local(F) and V_local(E), the following results were obtained:
[0099] Target object A Target object B: No conflict
[0100] Target object A Target object F: No conflict
[0101] Target object A Target object E: Conflicts exist in direction and spatial dependency breakage.
[0102] Furthermore, the system calculates the spatial structure consistency score: C_AB = 0.93, C_AF = 0.88, C_AE = 0.22. Since the system sets the consistency threshold to 0.70, therefore, A is retained. B, A F, and remove: A E.
[0103] Step 106: When a spatial structure conflict is detected, perform association correction processing on the conflicting target object according to the continuity of the propagation path, and generate a matching result corresponding to the target object to be matched based on the propagation candidate set after association correction.
[0104] The association correction process includes: removing candidate target objects whose propagation continuity is lower than a preset threshold, retaining candidate target objects that meet the spatial continuous propagation condition, and retaining candidate target objects that meet the spatial dependency continuity condition. Accordingly, the spatial propagation path corresponding to each candidate target object in the propagation candidate set can be obtained. Based on the dynamic propagation weight change trend between adjacent propagation nodes in the spatial propagation path, the propagation continuity of the corresponding spatial propagation path is determined, and the association correction process is performed on the candidate target objects according to the propagation continuity. Specifically, when the propagation continuity of the spatial propagation path is lower than the preset continuity threshold, the corresponding candidate target object is removed; when the spatial propagation path meets the spatial continuous propagation condition and the corresponding target object meets the spatial dependency continuity condition, the corresponding candidate target object is retained; and a final matching result corresponding to the target object to be matched is generated based on the association-corrected propagation candidate set.
[0105] For example, the system sets the continuity threshold to 0.70 and calculates the propagation path continuity: Path(A→B)=0.92, Path(A→F)=0.86, Path(A→E)=0.31; therefore, A is retained. B, A F, and remove: A E. Finally, the system generates the target matching result: target object A Target object B; Target object A Target object F.
[0106] This application's embodiments construct a spatial association graph and introduce dynamic propagation weights of distance, direction, and spatial attachment to achieve structured matching of multi-source sensing targets. By combining local spatial structure description and propagation consistency detection, the matching stability in complex scenarios can be improved. Through conflict correction and hierarchical compression, the false matching rate can be effectively reduced and the processing efficiency can be improved.
[0107] like Figure 5The diagram shown is a structural schematic of a multi-source sensing data matching system based on spatial correlation provided in an embodiment of this application, including:
[0108] The acquisition module 510 is used to acquire sensing data from multiple heterogeneous sensing sources, and to perform spatial mapping processing on the target objects detected by the sensing data of each heterogeneous sensing source based on a unified spatial reference system to form a target spatial distribution set; the multiple heterogeneous sensing devices include multiple devices among video sensing devices, positioning sensing devices, radio frequency identification devices and environmental sensing devices.
[0109] The construction module 520 is used to construct a spatial association graph based on the target spatial distribution set. The nodes in the spatial association graph represent target objects detected by various sensing data, and the edges in the spatial association graph represent spatial association relationships between different target objects. The spatial association relationships include adjacency relationships, directional association relationships, motion continuity relationships, and spatial dependence relationships.
[0110] Specifically, the construction module 520 is used to perform spatial neighborhood division for each target object in the target spatial distribution set, generating multiple local spatial sub-regions; based on the spatial positional relationships, movement direction relationships, and spatial dependency relationships between multiple target objects in each local spatial sub-region, a corresponding local association subgraph is constructed; each node in the local association subgraph represents a target object within the local spatial sub-region, and the edges in the local association subgraph represent the spatial association relationships between different target objects; according to the spatial change trends of two target objects connected to the same edge in the local association subgraph over multiple consecutive time periods, a corresponding dynamic propagation weight is assigned to the edge connecting the two target objects, and the dynamic propagation weight is... The propagation weight is used to characterize the spatial association stability between different target objects. According to the boundary overlap relationship between adjacent local spatial sub-regions, each local association subgraph is connected to generate a global spatial association graph. When there is a common target object or continuous spatial propagation path between adjacent local spatial sub-regions, cross-regional connection edges are established between the corresponding local association subgraphs. Based on the local spatial aggregation state of each target object, hierarchical association compression processing is performed on the global spatial association graph to generate a spatial association graph containing core association nodes and propagation association edges. The hierarchical association compression processing is used to associate and aggregate multiple target objects that are spatially continuous and have the same direction of movement, and to prune connection edges whose association stability is lower than a preset threshold.
[0111] Specifically, the spatial distance correlation parameter, motion direction correlation parameter, and spatial attachment correlation parameter between the two target objects can be calculated using the following formulas:
[0112]
[0113] Among them, R ij is the spatial distance correlation parameter between target object i and target object j, used to characterize their spatial proximity and stability in consecutive time intervals; T is the time window length, representing the total number of consecutive time intervals used for calculation; t is the time index, representing the t-th time interval within the time window length; d ij (t) represents the spatial distance between target object i and target object j at time t; Δd ij (t) represents the change in spatial distance between target object i and target object j between adjacent time points; σ is the distance decay coefficient, used to control the decay rate of the influence of spatial distance on the association strength, and the distance decay coefficient is determined based on the average spacing between target objects; η is the distance change stability coefficient, used to control the degree of influence of distance fluctuation on the association strength, and the distance change stability coefficient is determined based on the variance of distance change between target objects;
[0114]
[0115] Among them, V ij θ is the motion direction association parameter between target object i and target object j; T is the time window length, representing the total number of consecutive moments used for calculation; t is the time index, representing the t-th moment in the time window length; ij (t) represents the angle between the motion directions of target object i and target object j at time t; Δθ ij (t) represents the change in the angle between the motion directions of target object i and target object j between adjacent time points; κ is the direction stability decay coefficient, which is used to control the influence of direction changes on the association strength. The direction stability decay coefficient is determined based on the degree of dispersion of the motion directions between target objects.
[0116]
[0117] Among them, A ij is the spatial dependency parameter between target object i and target object j, used to characterize whether a stable spatial binding relationship exists between them; T is the time window length, representing the total number of consecutive moments used for calculation; t is the time index, representing the t-th moment in the time window length; d ij (t) represents the spatial distance between target object i and target object j at time t; Δr ij(t) represents the change in the relative position between target object i and target object j between adjacent time points, used to characterize spatial attachment stability; σ is the distance decay coefficient, used to control the decay rate of the influence of spatial distance on the association strength; μ is the relative position stability coefficient, used to control the influence of spatial structure stability, which is determined according to the magnitude of the relative position change between target objects;
[0118] Based on the spatial distance association parameters, motion direction association parameters, and spatial attachment association parameters between the two target objects, a corresponding dynamic propagation weight is assigned to the edge connecting the two target objects. The dynamic propagation weight is used to characterize the stability of the spatial association between different target objects, and the dynamic propagation weight is calculated based on the following formula:
[0119]
[0120] Among them, W ij R represents the dynamic propagation weights between target object i and target object j. ij V is the spatial distance correlation parameter between target object i and target object j. ij A is the motion direction association parameter between target object i and target object j. ij Let λ1, λ2, and λ3 be the spatial dependency parameters between target object i and target object j. Let λ1, λ2, and λ3 be the weight coefficients corresponding to the spatial distance parameter, motion direction parameter, and spatial dependency parameter, respectively. The values of λ1, λ2, and λ3 are all greater than zero and less than 1, and λ1+λ2+λ3=1.
[0121] In this embodiment, the values of the weighting coefficients λ1, λ2 and λ3 can be determined based on the spatial relationship sequence formed by the target object pair within a preset time window.
[0122] Specifically, a target object pair is formed by the first target object and the second target object. The spatial distance correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the spatial distance correlation parameter sequence is statistically analyzed to obtain a distance correlation stability index characterizing the stability of spatial distance correlation. The distance correlation stability index is then normalized to obtain a weighting coefficient λ1. Similarly, a target object pair is formed by the first target object and the second target object. The motion direction correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the motion direction correlation parameter sequence is statistically analyzed to obtain a direction correlation stability index characterizing the stability of motion direction correlation. The direction correlation stability index is then normalized to obtain a weighting coefficient λ2. Finally, a target object pair is formed by the first target object and the second target object. The spatial dependency correlation parameter sequence corresponding to the target object pair is calculated at consecutive time intervals, and the spatial dependency correlation parameter sequence is statistically analyzed to obtain a spatial dependency stability index characterizing the stability of spatial dependency correlation. The spatial dependency stability index is then normalized to obtain a weighting coefficient λ3.
[0123] The first generation module 530 is used to generate a local spatial association domain based on the spatial association graph for the target object to be matched corresponding to the first heterogeneous sensing source among the plurality of heterogeneous sensing sources. The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target direction distribution state, and the target spatial attachment state.
[0124] Specifically, the first generation module 530 is used to search for associated nodes in the spatial association graph with the target object to be matched as the central node, where the dynamic propagation weight between the target object and the target object is higher than a preset association threshold, and to form a local spatial association domain based on the associated nodes; according to the distribution state of the target objects in the local spatial association domain, to construct a target density distribution model, a target direction distribution model, and a target spatial dependency distribution model; the target density distribution model is used to characterize the degree of aggregation of target objects within the local space, the target direction distribution model is used to characterize the consistency of the motion directions between target objects, and the target spatial dependency distribution model is used to characterize the spatial binding stability between target objects; based on the target density distribution model, the target direction distribution model, and the target spatial dependency distribution model, to generate a local spatial structure description vector corresponding to the target object to be matched, so as to characterize the spatial structure state around the target object to be matched.
[0125] The second generation module 540 is used to perform association candidate propagation processing based on the local spatial association domain to generate a propagation candidate set. The association candidate propagation processing is used to extend and propagate the candidate target objects corresponding to the other heterogeneous sensing sources (excluding the first heterogeneous sensing source) along the spatial association graph according to the spatial association relationship. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path.
[0126] Specifically, the second generation module 540 is used to perform multi-level extended propagation along the propagation association edges in the spatial association graph, using the target object to be matched as the initial propagation node; during the propagation process, the dynamic propagation weights corresponding to each propagation association edge in the propagation path are accumulated to obtain the path propagation credibility; when the path propagation credibility is higher than the preset propagation credibility threshold, the corresponding candidate target object is retained as the propagation node to continue to perform extended propagation, and a propagation candidate set is generated according to the propagation path length, propagation level depth and path propagation credibility corresponding to each candidate target object.
[0127] The detection module 550 is used to perform spatial structure consistency detection on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is spatial structure conflict between target objects corresponding to different heterogeneous sensing sources. The spatial structure conflict includes target intersection conflict, movement direction conflict, space occupancy conflict and spatial dependency contradiction conflict.
[0128] Specifically, the detection module 550 is used to acquire each candidate target object in the propagation candidate set, and, taking each candidate target object as a central node, search for associated nodes in the spatial association graph whose dynamic propagation weight is higher than a preset association threshold to generate a local spatial association domain for the corresponding candidate target object; based on the target object distribution state in the local spatial association domain corresponding to each candidate target object, construct a target density distribution model, a target direction distribution model, and a target spatial dependency distribution model, and generate a local spatial structure description vector for the corresponding candidate target object based on the corresponding target density distribution model, target direction distribution model, and target spatial dependency distribution model; based on the local spatial structure description vector corresponding to each candidate target object, construct a spatial structure consistency constraint model to detect whether there is a spatial structure conflict between different candidate target objects; the spatial structure conflict includes: spatial position intersection conflict, movement direction reversal conflict, spatial occupation area overlap conflict, and spatial dependency relationship break conflict between target objects; and determine whether there is a spatial structure conflict between candidate target objects based on the spatial structure change trend of each candidate target object in continuous time.
[0129] Specifically, based on the spatial structure change trend of each candidate target object in continuous time, the corresponding spatial structure consistency score is calculated; when the spatial structure consistency score is lower than the preset consistency threshold, it is determined that there is a spatial structure conflict between the corresponding candidate target objects.
[0130] The correction module 560 is used to perform association correction processing on the conflict target object according to the continuity of the propagation path when a spatial structure conflict is detected, and generate a matching result corresponding to the target object to be matched based on the propagation candidate set after association correction.
[0131] The association correction process includes: removing candidate target objects whose propagation continuity is lower than a preset threshold, retaining candidate target objects that meet the spatial continuous propagation condition, and retaining candidate target objects that meet the spatial dependency continuity condition.
[0132] Specifically, the correction module 560 is used to obtain the spatial propagation path corresponding to each candidate target object in the propagation candidate set, determine the propagation continuity of the corresponding spatial propagation path based on the dynamic propagation weight change trend between adjacent propagation nodes in the spatial propagation path, and perform association correction processing on the candidate target objects according to the propagation continuity.
[0133] Specifically, the spatial propagation path corresponding to each candidate target object in the propagation candidate set is obtained. Based on the dynamic propagation weight change trend between adjacent propagation nodes in the spatial propagation path, the propagation continuity of the corresponding spatial propagation path is calculated, and the candidate target object is subjected to association correction processing according to the propagation continuity. Specifically, when the propagation continuity of the spatial propagation path is lower than a preset continuity threshold, the corresponding candidate target object is removed. When the spatial propagation path satisfies the spatial continuous propagation condition and the corresponding target object satisfies the spatial dependency continuity condition, the corresponding candidate target object is retained. The final matching result corresponding to the target object to be matched is generated based on the propagation candidate set after association correction.
[0134] This application's embodiments construct a spatial association graph and introduce dynamic propagation weights of distance, direction, and spatial attachment to achieve structured matching of multi-source sensing targets. By combining local spatial structure description and propagation consistency detection, the matching stability in complex scenarios can be improved. Through conflict correction and hierarchical compression, the false matching rate can be effectively reduced and the processing efficiency can be improved.
[0135] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the various processes of the above-described embodiments of the multi-source sensing data matching method based on spatial correlation, and achieves the same technical effects. To avoid repetition, it will not be described again here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.
[0136] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0137] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0138] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.
Claims
1. A multi-source sensing data matching method based on spatial correlation, characterized in that, Includes the following steps: The system acquires sensing data from multiple heterogeneous sensing sources, and performs spatial mapping processing on the target objects detected by the sensing data from each heterogeneous sensing source based on a unified spatial reference system to form a target spatial distribution set; the multiple heterogeneous sensing devices include multiple devices from video sensing devices, positioning sensing devices, radio frequency identification devices, and environmental sensing devices. A spatial association graph is constructed based on the target spatial distribution set. The nodes in the spatial association graph represent the target objects detected by each sensing data, and the edges in the spatial association graph represent the spatial association relationships between different target objects. The spatial association relationships include adjacency relationships, directional association relationships, motion continuity relationships, and spatial dependency relationships. For the target object to be matched corresponding to the first heterogeneous sensing source among the plurality of heterogeneous sensing sources, a local spatial association domain is generated based on the spatial association graph. The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target direction distribution state, and the target spatial attachment state. Based on the local spatial association domain, the association candidate propagation process is performed to generate a propagation candidate set. The association candidate propagation process is used to extend and propagate the candidate target objects corresponding to the heterogeneous sensing sources other than the first heterogeneous sensing source along the spatial association graph according to the spatial association relationship. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path. A spatial structure consistency detection is performed on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is a spatial structure conflict between target objects corresponding to different heterogeneous sensing sources. The spatial structure conflict includes target intersection conflict, movement direction conflict, space occupation conflict and spatial dependency contradiction conflict. When a spatial structure conflict is detected, the conflicting target object is subjected to association correction processing based on the continuity of the propagation path, and a matching result corresponding to the target object to be matched is generated based on the propagation candidate set after association correction.
2. The method according to claim 1, characterized in that, The construction of the spatial association graph based on the target spatial distribution set specifically includes: For each target object in the target spatial distribution set, spatial neighborhood division is performed to generate multiple local spatial sub-regions; Based on the spatial positional relationships, movement direction relationships, and spatial dependency relationships among multiple target objects in each local spatial sub-region, a corresponding local association subgraph is constructed; each node in the local association subgraph is used to represent each target object within the local spatial sub-region, and the edges in the local association subgraph are used to represent the spatial association relationships between different target objects. Based on the spatial change trends of two target objects connected to the same edge in the local association subgraph over multiple consecutive time periods, a corresponding dynamic propagation weight is assigned to the edge connecting the two target objects. The dynamic propagation weight is used to characterize the degree of spatial association stability between different target objects. Based on the boundary overlap relationship between adjacent local spatial sub-regions, each local associated subgraph is connected to generate a global spatial association graph. When there is a common target object or continuous spatial propagation path between adjacent local spatial sub-regions, cross-regional connection edges are established between the corresponding local associated subgraphs. Based on the local spatial aggregation state of each target object, a hierarchical association compression process is performed on the global spatial association graph to generate a spatial association graph containing core association nodes and propagation association edges. The hierarchical association compression process is used to associate and aggregate multiple target objects that are spatially continuous and have the same direction of movement, and to prune connection edges whose association stability is lower than a preset threshold.
3. The method according to claim 2, characterized in that, The generation of local spatial association domains based on the spatial association graph specifically includes: Using the target object to be matched as the central node, search the spatial association graph for association nodes whose dynamic propagation weight between them and the target object is higher than a preset association threshold, and form a local spatial association domain based on the association nodes; Based on the distribution state of target objects in the local spatial association domain, a target density distribution model, a target direction distribution model, and a target spatial attachment distribution model are constructed. The target density distribution model is used to characterize the degree of aggregation of target objects within the local spatial range, the target direction distribution model is used to characterize the degree of consistency of the motion directions between target objects, and the target spatial attachment distribution model is used to characterize the degree of spatial binding stability between target objects. Based on the target density distribution model, the target orientation distribution model, and the target spatial dependence distribution model, a local spatial structure description vector corresponding to the target object to be matched is generated to characterize the spatial structure state around the target object to be matched.
4. The method according to claim 1, characterized in that, The step of performing association candidate propagation processing based on the local spatial association domain to generate a propagation candidate set specifically includes: Using the target object to be matched as the initial propagation node, multi-level extended propagation is performed along the propagation association edges in the spatial association graph; During the propagation process, the dynamic propagation weights corresponding to each propagation-related edge in the propagation path are accumulated to obtain the path propagation credibility. When the path propagation credibility is higher than the preset propagation credibility threshold, the corresponding candidate target object is retained as a propagation node to continue the extended propagation, and a propagation candidate set is generated according to the propagation path length, propagation level depth and path propagation credibility of each candidate target object.
5. The method according to claim 1, characterized in that, The step of performing spatial structure consistency detection on the propagation candidate set specifically includes: Each candidate target object in the propagation candidate set is obtained. Taking each candidate target object as the central node, the association node with the dynamic propagation weight between it and the corresponding candidate target object is higher than the preset association threshold in the spatial association graph, so as to generate the local spatial association domain of the corresponding candidate target object. Based on the distribution state of target objects in the local spatial association domain corresponding to each candidate target object, construct a target density distribution model, a target direction distribution model, and a target spatial dependency distribution model. Based on the corresponding target density distribution model, target direction distribution model, and target spatial dependency distribution model, generate the local spatial structure description vector of the corresponding candidate target object. Based on the local spatial structure description vectors corresponding to each candidate target object, a spatial structure consistency constraint model is constructed to detect whether there are spatial structure conflicts between different candidate target objects. The spatial structure conflicts include: spatial position intersection conflict, movement direction reversal conflict, spatial occupation area overlap conflict, and spatial dependency relationship break conflict between target objects. The process of performing association correction on conflicting target objects based on the continuity of the propagation path specifically includes: Obtain the spatial propagation path corresponding to each candidate target object in the propagation candidate set. Based on the dynamic propagation weight change trend between adjacent propagation nodes in the spatial propagation path, determine the propagation continuity of the corresponding spatial propagation path, and perform association correction processing on the candidate target objects according to the propagation continuity.
6. A multi-source sensing data matching system based on spatial correlation, characterized in that, include: The acquisition module is used to acquire sensing data from multiple heterogeneous sensing sources, and perform spatial mapping processing on the target objects detected by the sensing data of each heterogeneous sensing source based on a unified spatial reference system to form a target spatial distribution set; the multiple heterogeneous sensing devices include multiple devices from video sensing devices, positioning sensing devices, radio frequency identification devices and environmental sensing devices; The construction module is used to construct a spatial association graph based on the target spatial distribution set. The nodes in the spatial association graph represent target objects detected by each sensing data, and the edges in the spatial association graph represent spatial association relationships between different target objects. The spatial association relationships include adjacency relationships, directional association relationships, motion continuity relationships, and spatial dependency relationships. The first generation module is used to generate a local spatial association domain based on the spatial association graph for the target object to be matched corresponding to the first heterogeneous sensing source among the plurality of heterogeneous sensing sources. The local spatial association domain is used to characterize the spatial structure state within a preset range around the target object to be matched. The spatial structure state includes the target density distribution state, the target direction distribution state, and the target spatial attachment state. The second generation module is used to perform association candidate propagation processing based on the local spatial association domain to generate a propagation candidate set. The association candidate propagation processing is used to extend and propagate the candidate target objects corresponding to the heterogeneous sensing sources other than the first heterogeneous sensing source along the spatial association graph according to the spatial association relationship. The candidate target objects in the propagation candidate set establish association relationships through the spatial propagation path. The detection module is used to perform spatial structure consistency detection on the propagation candidate set. The spatial structure consistency detection is used to detect whether there is spatial structure conflict between target objects corresponding to different heterogeneous sensing sources. The spatial structure conflict includes target intersection conflict, movement direction conflict, space occupancy conflict and spatial dependency contradiction conflict. The correction module is used to perform association correction processing on the conflicting target object according to the continuity of the propagation path when a spatial structure conflict is detected, and to generate a matching result corresponding to the target object to be matched based on the propagation candidate set after association correction.
7. The system according to claim 6, characterized in that, The construction module is specifically used to divide the spatial neighborhood of each target object in the target spatial distribution set, generating multiple local spatial sub-regions; based on the spatial positional relationship, movement direction relationship, and spatial dependency relationship between multiple target objects in each local spatial sub-region, a corresponding local association subgraph is constructed; each node in the local association subgraph represents each target object in the local spatial sub-region, and the edges in the local association subgraph represent the spatial association relationship between different target objects; according to the spatial change trend of two target objects connected by the same edge in the local association subgraph over multiple consecutive time periods, a corresponding dynamic propagation weight is assigned to the edge connecting the two target objects, and the dynamic propagation weight is used to characterize the degree of spatial association stability between different target objects; Based on the boundary overlap relationship between adjacent local spatial sub-regions, each local associated subgraph is connected to generate a global spatial association graph. When there is a common target object or continuous spatial propagation path between adjacent local spatial sub-regions, cross-regional connection edges are established between the corresponding local associated subgraphs. Based on the local spatial aggregation state of each target object, a hierarchical association compression process is performed on the global spatial association graph to generate a spatial association graph containing core association nodes and propagation association edges. The hierarchical association compression process is used to associate and aggregate multiple target objects that are spatially continuous and have the same direction of movement, and to prune connection edges whose association stability is lower than a preset threshold.
8. The system according to claim 7, characterized in that, The first generation module is specifically used to use the target object to be matched as the central node, search for associated nodes in the spatial association graph whose dynamic propagation weight between the target object and the target object is higher than a preset association threshold, and form a local spatial association domain based on the associated nodes; according to the distribution state of the target objects in the local spatial association domain, construct a target density distribution model, a target direction distribution model, and a target spatial attachment distribution model; the target density distribution model is used to characterize the degree of aggregation of target objects within the local space, the target direction distribution model is used to characterize the degree of consistency of the movement directions between target objects, and the target spatial attachment distribution model is used to characterize the degree of spatial binding stability between target objects; Based on the target density distribution model, the target orientation distribution model, and the target spatial dependence distribution model, a local spatial structure description vector corresponding to the target object to be matched is generated to characterize the spatial structure state around the target object to be matched.
9. The system according to claim 6, characterized in that, The second generation module is specifically used to perform multi-level extended propagation along the propagation association edges in the spatial association graph, using the target object to be matched as the initial propagation node; during the propagation process, the dynamic propagation weights corresponding to each propagation association edge in the propagation path are accumulated to obtain the path propagation credibility; when the path propagation credibility is higher than the preset propagation credibility threshold, the corresponding candidate target object is retained as a propagation node to continue to perform extended propagation, and a propagation candidate set is generated according to the propagation path length, propagation level depth and path propagation credibility corresponding to each candidate target object.
10. The system according to claim 6, characterized in that, The detection module is specifically used to acquire each candidate target object in the propagation candidate set, and, taking each candidate target object as a central node, search for associated nodes in the spatial association graph whose dynamic propagation weight is higher than a preset association threshold to generate a local spatial association domain for the corresponding candidate target object; based on the target object distribution state in the local spatial association domain corresponding to each candidate target object, construct a target density distribution model, a target direction distribution model, and a target spatial dependency distribution model, and generate a local spatial structure description vector for the corresponding candidate target object based on the corresponding target density distribution model, target direction distribution model, and target spatial dependency distribution model; and construct a spatial structure consistency constraint model based on the local spatial structure description vector corresponding to each candidate target object to detect whether there is a spatial structure conflict between different candidate target objects. The spatial structural conflicts include: conflicts of spatial position intersection between target objects, conflicts of opposite movement directions, conflicts of overlapping spatial occupied areas, and conflicts of broken spatial attachment relationships. The correction module is specifically used to obtain the spatial propagation path corresponding to each candidate target object in the propagation candidate set, determine the propagation continuity of the corresponding spatial propagation path based on the dynamic propagation weight change trend between adjacent propagation nodes in the spatial propagation path, and perform association correction processing on the candidate target objects according to the propagation continuity.
