A Big Data Compression Method for Mobile Object Trajectories Oriented to Indoor Scene Semantics

Abstract

The present invention discloses a method for compressing big data of mobile object trajectories oriented to indoor scene semantics, including: obtaining indoor scene data, modeling important spatial entity elements in the scenes within a floor and across floors to obtain topological components of indoor scene elements; modeling indoor pedestrian sub-trajectories and indoor geographical space elements in sequence, respectively establishing local overlapping semantic models for sub-trajectory - local corridor elements and sub-trajectory - regions to obtain their semantic relationships; introducing an encoding method to obtain overlapping semantic information between sub-trajectory tuples and indoor local corridors and regions; reconstructing the sub-trajectory semantics into semantic trajectories for different indoor geographical space elements, while compressing and segmenting the original trajectories, performing a direct spatial connection between the trajectory data and indoor scene elements to enhance the expression ability of the trajectories.

Description

Technical Field

[0001] The present invention belongs to the technical field of data compression, and in particular relates to a mobile object trajectory big data compression method oriented to indoor scene semantics. Background Art

[0002] In recent years, in the context of smart cities, complex buildings represented by large commercial centers, underground parking lots, apartments, office buildings, residences and large indoor public places are increasing in cities. At the same time, with the increasing popularity of mobile smart terminal devices and the continuous development of indoor positioning technology, people's exploration and research of the real world has gradually moved from outdoor space to indoor space. The demand for indoor location services (LBS, Location Based Services) such as indoor navigation, path planning, and emergency evacuation is also increasing. Research and application related to indoor positioning trajectories are increasingly valued by researchers. How to store and manage these large amounts of data is a difficult task, so trajectory compression has received more and more attention in recent years. Trajectory compression can solve problems such as huge data volume, increased query latency, and data redundancy. Therefore, trajectory compression has become an urgent need to store and manage massive trajectory data, and can provide better services.

[0003] The existing trajectory compression methods and semantic trajectory compression models are largely targeted at outdoor environments. The main defects are: (1) The current mainstream human mobile trajectory compression models are more targeted at outdoor environments, lacking in-depth research on complex trajectories in indoor environments, and cannot effectively support people's needs for indoor location services such as indoor navigation, path planning, and emergency evacuation. (2) The current trajectory compression methods cannot well establish and utilize the direct relationship between mobile objects and surrounding scenes, so that simply analyzing and studying the original trajectory compression data is difficult to reveal the indoor user's movement intention. (3) The current mainstream trajectory compression methods lack the combination of indoor network topology and context information, and cannot coordinately analyze the topological behavior semantics of mobile objects and indoor environment semantics. It is difficult to use intuitive data organization methods to manage the spatiotemporal topological semantics of trajectory data, resulting in the specific goal of analyzing the trajectory of mobile objects in indoor scenes being difficult to achieve, and it is impossible to refine the complex movement behavior of indoor mobile objects. (4) The current trajectory compression methods do not consider the specific analysis of the leap-forward movement behavior of mobile objects between multiple floors in indoor scenes, and cannot obtain the complex movement behavior of mobile objects in a complete indoor space.

[0004] As people spend most of their daily lives indoors, and the scale of large indoor public places continues to grow, traditional trajectory compression algorithms are facing bottlenecks and challenges, and it is necessary to develop new trajectory compression algorithms that can meet the needs of indoor environments. In the context of smart cities, new requirements are also put forward, and it is necessary to consider indoor space information and semantic information for analysis and integration, reasonably model and manage indoor trajectories, and make indoor space data models reasonably and effectively expressed. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a mobile object trajectory big data compression method for indoor scene semantics in view of the deficiencies of the prior art. In order to compress massive indoor trajectory data, a trajectory topological semantic model under indoor space constraints is constructed, and indoor trajectories are semantically encoded. In terms of spatiotemporal topological semantic management, the trajectory data is reasonably stored and expressed with intuitive data, and the specific analysis of the leap-forward movement behavior of mobile objects between multiple floors is realized, and the complex movement behavior of mobile objects in a complete indoor space and the multi-level semantic information of trajectories are refined.

[0006] To solve the above technical problems, the present invention includes:

[0007] A method for compressing big data of moving object trajectories for indoor scene semantics comprises the following steps:

[0008] S1. Acquire indoor scene data, wherein the indoor scene includes a floor scene and a cross-floor scene, model important spatial entity elements in the floor scene and the cross-floor scene, and obtain the topological components of the indoor scene elements;

[0009] S2. A scene topology tuple established by using a moving object trajectory topology tuple and a topological component of an indoor scene element is used to construct a local overlapping semantic model of an indoor scene, so as to capture the dynamic behavior semantics of the trajectory in the indoor scene element; the moving object trajectory topology tuple is composed of a front sub-trajectory, a key point and a rear sub-trajectory; the indoor scene element includes an indoor local corridor element and an indoor area element, and the indoor scene local overlapping semantic model includes a sub-trajectory tuple-indoor local corridor element overlapping semantic model and a sub-trajectory tuple-indoor area element overlapping semantic model constructed by using three topological invariants in the moving object trajectory topology tuple;

[0010] S3. Encode the overlapping semantics of the sub-trajectory tuple and the indoor scene elements using a character-based encoding method to obtain the effective overlapping semantics of the sub-trajectory for the local overlapping semantic model of the indoor scene;

[0011] S4. Reconstruct the trajectory semantics according to the local overlapping semantic model of the indoor scene, organize the semantic information in the model according to the time dimension in an encoding manner, and compress the original trajectory into a complete semantic trajectory;

[0012] S5. For the semantic trajectory of the moving object reconstructed in step S4, construct a semantic parser for the compressed trajectory. Based on the effective overlapping semantics of the sub-trajectory - local corridor elements and sub-trajectory - indoor area elements obtained in step S3, perform semantic parsing on the pedestrian trajectory to obtain a semantic trajectory managed and organized in units of local semantics.

[0013] Furthermore, in step S1, perform data preprocessing on the indoor scene data, including format conversion and coordinate conversion.

[0014] Furthermore, in step S1, for the in-floor scene, obtain the attribute information of the POI areas, corridors, and public areas on each floor. For the cross-floor scene, obtain the attribute information of the vertical elevators, escalators, and elevators on each floor to obtain the spatio-temporal information of the indoor scene for semantic trajectory query.

[0015] Furthermore, in step S2, the moving object trajectory consists of sub-trajectories and key points. The key points are the positions where the topological relationship changes during the movement of the pedestrian. The sub-trajectory earlier than the key point time attribute in the key point neighborhood is the pre-sub-trajectory, and the sub-trajectory later than the key point time attribute is the post-sub-trajectory.

[0016] Furthermore, in step S2, the indoor local corridor consists of multiple corridors in the indoor venue. Both ends of each corridor are used as its nodes, and the nodes are used as the boundaries of each corridor. The arc segment between the boundaries is used as the inside of the corridor, and the topological space outside the boundaries is used as the outside of the corridor. The topological components of the indoor local corridor elements are composed of the outside, boundaries, and inside; the topological components of the indoor area elements are composed of the front, back, outside, and boundaries. The front is the low-floor area of the same scene, the back is the high-floor area of the same scene, and the outside and boundaries are the boundaries and outside of the corresponding areas of the front and back, respectively.

[0017] Furthermore, in step S2, the semantic calculation of the overlapping of each topological component in the sub-trajectory tuple with the topological components in the indoor local corridor elements is as follows:

[0018]

[0019] In the formula, the three topological invariants in the sub-trajectory tuple are respectively encoded as the pre-sub-trajectory IT pre 、the key point IT key 、the post-sub-trajectory IT post, L1 and L2 represent nodes 1 and 2 at both ends of the corridor, respectively, and L0 represents the interior of the corridor except the nodes; the default positive direction of movement on the corridor is from node 1 to node 2, and vice versa; each value in the matrix represents the overlap of the sub-trajectory topological invariant and the corridor element topological invariant, and its value is empty or 1, -1, 0. Empty means that there is an empty set in the two types of topological invariants, 1 and -1 respectively indicate that there is a positive intersection and a reverse intersection between the two types of topological invariants, and 0 indicates that there is no intersection between the two types of topological invariants;

[0020] The semantics of the overlap calculation of each topological component in the sub-trajectory tuple and the topological component in the indoor area element is as follows:

[0021]

[0022] In the formula, DR d ,DR u ,DR o and DR - They respectively represent the front of the same scene, i.e. the low-floor area, the back of the same scene, i.e. the high-floor area, the boundary of the same scene, and the external area of ​​the same scene; each value in the matrix represents the overlap of the sub-trajectory topological invariant and the topological invariant of the indoor area elements, and its value can be empty or 0 or 1. Empty means that there is an empty set in the two types of topological invariants, 0 means that there is no intersection between the two types of topological invariants, and 1 means that there is an intersection between the two types of topological invariants.

[0023] Furthermore, in step S3, the encoding method of the sub-trajectory tuple and the indoor local corridor element includes the valid overlapping semantics in the 9 overlapping parts, and the lowercase letters i, b, and e are used to represent the internal overlap of the moving object and the corridor, the boundary overlap of the moving object and the corridor, and the external overlap of the moving object and the corridor respectively; the node numbers b1 and b2 are used to distinguish the movement direction of the sub-trajectory in the corridor, so far i represents the attribute of the intersection of the sub-trajectory and the corridor from the node b1 to b2, and -i represents the attribute of the intersection of the sub-trajectory and the corridor from the node b2 to b1; if there is no null in the two types of topological invariants, the non-null corresponding mark is i, -i, b1, b2, e, otherwise it is marked as null; the encoding of the sub-trajectory tuple and the indoor local corridor element is marked in the form of xyz, and x, y, and z respectively represent the topological overlapping relationship between the front sub-trajectory, the key point, the rear sub-trajectory and the corridor element;

[0024] The encoding method of the sub-trajectory tuple and the indoor area element includes the valid overlapping semantics in 16 overlapping parts, and uses the capital letters I, B, and E to represent the internal overlap between the moving object and the indoor area element, the boundary overlap between the moving object and the indoor area element, and the external overlap between the moving object and the indoor area element respectively; uses the internal area numbers I1 and I2 to distinguish the sub-trajectories inside the low-floor area and the high-floor area in the same scene. If there is no empty in the two types of topological invariants, the non-empty ones are correspondingly marked as I1, I2, B, and E, otherwise they are marked as empty; the encoding of the sub-trajectory tuple and the indoor area element is marked in the form of XYZ, and X, Y, and Z respectively represent the topological overlap relationships between the previous sub-trajectory, the key point, the subsequent sub-trajectory and the indoor area element.

[0025] Further, in the step S4, the indoor scene local overlapping semantic model compresses the moving trajectory in the same type of encoding method for the indoor local corridor elements and the indoor area elements. The general expression form of the indoor scene moving trajectory is:

[0026]

[0027] Among them, Tcode represents the semantic encoding of the sub-trajectory with different elements in the indoor scene, "@" is the connection symbol connecting the indoor scene elements, "Q" is the unique number corresponding to the sub-trajectory tuple for a certain indoor scene element, " / " is the connection symbol connecting the indoor floor numbers, "F" is the unique number representing the floor where the topological component of the key point of the sub-trajectory is located, "*" is the connection symbol connecting the moving modes of the moving object, and "W" is the moving mode adopted by the moving object; use "+" to connect the different semantics of the topological tuple where the key point of the sub-trajectory is located, and use "↓" to connect the combined semantics of the floor change between the sub-trajectories among the indoor area elements;

[0028] Reconstructing the trajectory semantics includes a semantic trajectory reconstruction mode based on indoor local corridor elements and a semantic trajectory reconstruction mode based on indoor area elements. Their trajectory formulas are respectively:

[0029]

[0030]

[0031] The beneficial effects of the present invention are:

[0032] (1) The local overlapping semantic model of indoor scenes constructed by the present invention highlights the spatial constraints of mobile behavior in indoor scenes, adds the geographical semantics of mobile trajectories to the restrictions of corridor elements and regional elements in indoor scenes, and studies the mobile behavior characteristics of mobile trajectories for indoor scene elements. Focusing on the characteristics of mobile trajectories from the time dimension and space dimension, topological information is added to integrate indoor semantic information such as corridor elements and POI areas in indoor places with mobile trajectories, so as to analyze the mobile behavior of objects in indoor places and meet the needs of indoor space location services such as topic query and analysis; (2) The trajectory semantic coding and compression method proposed by the present invention highlights the expression of the semantics of mobile behavior of mobile objects in indoor places, effectively compresses massive indoor trajectories, and reconstructs the behavioral relationship of trajectories for different geographical elements in indoor scenes through local semantics of indoor scenes. Each semantic is encoded and reconstructed to enhance the expression ability of the original trajectory, so as to facilitate the semantic query of indoor mobile object trajectories; (3) The local overlapping semantic model of indoor scenes highlights the semantic analysis of cross-floor trajectories in indoor scenes. The semantic information of mobile devices used by mobile objects to move across floors is added to the general formula for reconstructing trajectory semantics, thereby refining the complex movement behavior of mobile objects in the complete indoor space; (4) Topological semantics are used to perform spatiotemporal queries on mobile trajectories, and specific scenario analysis is proposed for different indoor location service needs to explain people's activity behaviors and explore the potential value of mobile objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flowchart for constructing a local overlapping semantic model of indoor scenes to compress indoor trajectories;

[0034] Figure 2 It is an example diagram of trajectory and semantic trajectory reconstruction of indoor local corridor elements;

[0035] Figure 3 It is an example graph of trajectory and semantic trajectory reconstruction of indoor local area elements;

[0036] Figure 4 The query visualization diagram is "Which mobile objects have been to "SexyTea" (POI25) during the period of 2022 / 06 / 15 13:00 to 2022 / 06 / 15 15:00?"

[0037] Figure 5 The query visualization diagram is "When did mobile object 5 enter the "Ten O'Clock Bookstore" (POI35) area?"

[0038] Figures 6 - 12It is a visualization graph for querying the question: "During the period from 14:00 on June 15, 2022 to 15:30 on June 15, 2022, which moving objects took the vertical elevator downstairs?" Detailed implementation manners

[0039] To facilitate the understanding of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners. Those skilled in the art should understand that the embodiments are only for helping to understand the present invention and should not be regarded as specific limitations on the present invention.

[0040] As Figure 1 shown, the present invention provides a method for compressing big data of moving object trajectories for indoor scene semantics, including the following steps:

[0041] Step S1. Obtain indoor scene data. The indoor scene includes in-floor scenes and cross-floor scenes. Model important spatial entity elements in the in-floor scenes and cross-floor scenes to obtain topological components of indoor scene elements.

[0042] The indoor scene data covers all the data available in indoor scenes. The important spatial entity elements are surface elements (for example, a room on a floor) and line elements (for example, a section of corridor on a floor) extracted from indoor scenes. These extracted spatial entity elements are used to construct a local overlapping semantic model in step S2. Decompose the abstracted scene elements (line elements and surface elements) into a combination of 3 topological components, that is, the geometric description of each element includes three topological components: internal, external, and boundary.

[0043] Extract topological components from trajectories and indoor scene elements respectively based on the method of point-set topology theory. When modeling, the trajectory is regarded as a special directed line element. The three obtained topological components are expressed and stored in the form of a tuple, so it is also called a topological tuple; for a line element, the two endpoints of the line are the "boundary" in the topological components of the line element, the line body between the two endpoints is the "internal" in the topological components of the line element, and the rest is the "external" in the topological components of the line element; for a surface element, the boundary line of the surface is the "boundary" in the topological components of the surface element, the space inside the surface is the "internal" in the topological components of the surface element, and the rest is the "external" in the topological components of the surface element.

[0044] Step S2. Construct an indoor scene local overlapping semantic model by using the scene topology tuple established from the topological tuple of the moving object trajectory and the topological components of the indoor scene elements to capture the dynamic behavior semantics of the trajectories in the indoor scene elements; the topological tuple of the moving object trajectory consists of a front sub-trajectory, a key point, and a rear sub-trajectory; the indoor scene elements include indoor local corridor elements and indoor area elements, and the indoor scene local overlapping semantic model includes a sub-trajectory tuple - indoor local corridor element overlapping semantic model and a sub-trajectory tuple - indoor area element overlapping semantic model constructed by using three topological invariants in the topological tuple of the moving object trajectory.

[0045] Step S3. Use an encoding method composed of characters to encode the overlapping semantics between the sub-trajectory tuple and the indoor scene elements to obtain the effective overlapping semantics of the sub-trajectory for the indoor scene local overlapping semantic model.

[0046] Step S4. According to the indoor scene local overlapping semantic model, reconstruct the trajectory semantics, organize the semantic information in the model in the time dimension in an encoding manner, and compress the original trajectory into a complete semantic trajectory to facilitate the management and storage of the local behavior topological semantic information of the indoor scene moving trajectory.

[0047] In Step S1, for the indoor scene data (indoor venue location information) provided by Amap or Baidu, data preprocessing operations such as format conversion and coordinate conversion should be performed on the data to meet the standards of the multi-source geographic spatio-temporal information modeling. In terms of coordinate conversion, the WGS-84 coordinate system is mainly considered as the mainstream coordinate system, and the geographical locations in the multi-source data are uniformly converted to WGS-84.

[0048] In Step S1, process and store the indoor scene data on the ArcGIS Pro platform, and model the important spatial entity elements in the intra-floor and cross-floor scenes. For the intra-floor scene, obtain the attribute information of each POI area, corridor, and public area within each floor. For the cross-floor scene, obtain the attribute information of each vertical elevator, escalator, and elevator within each floor to obtain the indoor scene spatio-temporal information for semantic trajectory query.

[0049] In step S2, the moving object trajectory consists of sub-trajectories and key points. The key points are the positions where the topological relationship changes during the movement of pedestrians. The sub-trajectory that occurs earlier than the key point in the neighborhood of the key point is the front sub-trajectory, and the sub-trajectory that occurs later than the key point is the rear sub-trajectory. The indoor local corridor consists of multiple corridors in the indoor area. The two ends of each corridor are used as its nodes, the nodes are used as the boundaries of each corridor, the arc segments between the boundaries are used as the interior of the corridor, and the topological space outside the boundary is used as the outside of the corridor. The topological components of the indoor local corridor elements are composed of the outside, the boundary, and the interior; the topological components of the indoor area elements are composed of the front, the back, the outside, and the boundary. The front is the low-floor area of the same scene, the back is the high-floor area of the same scene, and the outside and the boundary are the boundaries and the outside of the areas corresponding to the front and the back, respectively.

[0050] In step S2, for the sub-trajectory tuple - indoor local corridor element overlapping semantic model. Use the three topological invariants in the trajectory tuple to construct the local overlapping semantic model of the sub-trajectory tuple and the indoor local corridor element, and obtain the topological semantics of different sub-trajectories relative to the indoor local corridor element in the time dimension. The semantic situation of the overlapping calculation of each topological component in the sub-trajectory tuple and the topological component in the indoor local corridor element is shown in formula (1):

[0051]

[0052] In formula (1), the three topological invariants in the sub-trajectory tuple are respectively encoded as the front sub-trajectory IT pre , the key point IT key , the rear sub-trajectory IT post , L1 and L2 respectively represent node 1 and node 2 at both ends of the corridor, and L0 represents the interior of the corridor except the nodes; among them, the default positive direction of movement on the corridor is from node 1 to node 2, and the opposite is the reverse direction; each value in the matrix represents the overlapping situation of the sub-trajectory topological invariant and the corridor element topological invariant, and its value is empty or 1, -1, 0. Empty means there is an empty set in the two types of topological invariants, 1 and -1 respectively represent the positive intersection and the reverse intersection between the two types of topological invariants, and 0 means there is no intersection between the two types of topological invariants.

[0053] For the sub-trajectory tuple - indoor area element overlapping semantic model. Use the three topological invariants in the trajectory tuple to construct the local overlapping semantic model of the sub-trajectory tuple and the indoor area element, and obtain the topological semantics of different sub-trajectories relative to the indoor area element in the time dimension. The semantic situation of the overlapping calculation of each topological component in the sub-trajectory tuple and the topological component in the indoor area element is shown in formula (2):

[0054]

[0055] In formula (2), DRd , DR u , DR o and DR - respectively represent the front, i.e., the low - floor area, the back, i.e., the high - floor area, the boundary, and the external area of the same scene; each value in the matrix represents the overlapping situation between the topological invariant of the sub - trajectory and the topological invariant of the indoor area element, and its value can be empty or 0, 1. Empty means there is an empty set in one of the two types of topological invariants, 0 means there is no intersection between the two types of topological invariants, and 1 means there is an intersection between the two types of topological invariants. Among them, if there is no empty value in the topological invariant of the sub - trajectory, the overlapping situation between the key points of the sub - trajectory and the front and back of the indoor area element is determined by its corresponding post - sub - trajectory.

[0056] In step S3, an encoding method composed of characters is introduced to reasonably store and understand the overlapping semantic information between the sub - trajectory tuple and the local elements of the indoor scene. Characters are used as substitute symbols for semantics to record the effective overlapping semantics of the sub - trajectory for the local overlapping semantic model of the indoor scene.

[0057] For the encoding method of the sub - trajectory tuple and the local indoor corridor element, it includes the effective overlapping semantics in 9 overlapping parts. The lowercase letters i, b, and e are used to represent the internal overlap of the moving object and the corridor, the boundary overlap of the moving object and the corridor, and the external overlap of the moving object and the corridor respectively; the node numbers b1 and b2 are used to distinguish the moving direction of the sub - trajectory on the corridor. Thus, i represents that the attribute of the sub - trajectory intersecting the corridor is from node b1 to b2, and - i represents that the attribute of the sub - trajectory intersecting the corridor is from node b2 to b1; if there is no empty in the two types of topological invariants, the non - empty ones are marked as i, - i, b1, b2, e respectively, otherwise marked as empty; the encoding of the sub - trajectory tuple and the local indoor corridor element is marked in the form of xyz, where x, y, and z respectively represent the topological overlapping relationships between the pre - sub - trajectory, the key point, and the post - sub - trajectory and the corridor element.

[0058] For the encoding method of the sub - trajectory tuple and the indoor area element, it includes the effective overlapping semantics in 16 overlapping parts. The capital letters I, B, and E are used to represent the internal overlap of the moving object and the indoor area element, the boundary overlap of the moving object and the indoor area element, and the external overlap of the moving object and the indoor area element respectively; the internal numbers I1 and I2 of the area are used to distinguish the inside of the low - floor area and the inside of the high - floor area of the sub - trajectory in the same scene. If there is no empty in the two types of topological invariants, the non - empty ones are marked as I1, I2, B, E respectively, otherwise marked as empty; the encoding of the sub - trajectory tuple and the indoor area element is marked in the form of XYZ, where X, Y, and Z respectively represent the topological overlapping relationships between the pre - sub - trajectory, the key point, and the post - sub - trajectory and the indoor area element.

[0059] In step S4, the indoor scene local overlapping semantic model compresses the movement trajectory in the same type of coding manner for the indoor local corridor elements and indoor area elements. The general expression form of the indoor scene movement trajectory is shown in formula (3):

[0060]

[0061] In formula (3), Tcode represents the semantic coding of the sub-trajectories with different elements of the indoor scene. "@" is the connection symbol for connecting the indoor scene elements. "Q" is the unique number corresponding to a certain indoor scene element for the sub-trajectory tuple. " / " is the connection symbol for connecting the indoor floor numbers. "F" is the unique number representing the floor where the topological component of the key point of the sub-trajectory is located. "*" is the connection symbol for connecting the movement modes of the moving object. "W" is the movement mode adopted by the moving object. Since the semantics of the indoor local elements corresponding to the same sub-trajectory tuple are different, "+" is used to connect the different semantics of the topological tuples where the key points of the sub-trajectory are located. If the inside of the sub-trajectory intersects with the front and back of the indoor scene area, the sub-trajectory must pass through the floor. Therefore, the trajectory between the front and back of the same area is not considered, and the combined semantics of the floor change of the sub-trajectory between the indoor area elements are directly connected by "↓".

[0062] The reconstruction of the trajectory semantics includes the semantic trajectory reconstruction mode based on the indoor local corridor elements (T-L) and the semantic trajectory reconstruction mode based on the indoor area elements (T-R). Their trajectory formulas are shown in formulas (4) and (5) respectively:

[0063]

[0064]

[0065] In the semantic trajectory reconstruction mode based on the indoor local corridor elements, the behavioral topological semantics are encoded in the form of xyz, and "Q" is the unique number corresponding to the indoor local corridor for the sub-trajectory tuple. Formula (4) is the T-L trajectory formula, and a specific example is Figure 2 as shown. Based on the T-L trajectory formula, the topological semantic reconstruction of the trajectories having topological relationships with the indoor corridors f, e, g, h, j, i is carried out according to the time dimension.

[0066] In the semantic trajectory reconstruction mode based on the indoor area elements, the behavioral topological semantics are encoded in the form of XYZ, and "Q" is the unique number corresponding to a certain indoor area for the sub-trajectory tuple. Formula (5) is the T-R trajectory formula, and a specific example is Figure 3 as shown. Based on the T-R trajectory formula, the topological semantic reconstruction of the trajectories having topological relationships with the indoor areas numbered 8, 25, 33 is carried out according to the time dimension.

[0067] The above formula consists of multiple tuples in the complete trajectory of a moving object. Each tuple represents the topological semantics of key points in different indoor local elements, and different semantics are arranged in chronological order. Figure 2 and Figure 3 In [Figure 1], the solid line with an arrow represents the trajectory of a moving object. The points in the solid line with an arrow are the key points where the topological state changes. The dashed line represents the local corridor of the indoor scene, and the points at both ends of the dashed line are the nodes at both ends of the corridor. The numbers are identifiers representing the numbers of indoor POI areas, the capital letters are identifiers representing the numbers of indoor public areas, and the lowercase letters are identifiers representing the numbers of indoor corridor elements. To facilitate the identification of moving devices such as stairs, escalators, and elevators, corresponding icons are used for identification.

[0068] Finally, experiments are conducted using pedestrian indoor movement trajectory data. The semantic trajectory is reconstructed according to the local overlapping semantic model, and the generated indoor semantic trajectory is queried and retrieved to prove the effectiveness and feasibility of the model.

[0069] For the semantic trajectory of the moving object reconstructed in step S4, a semantic parser for the compressed trajectory is constructed. Based on the effective overlapping semantics of the sub-trajectory - local corridor elements and sub-trajectory - indoor area elements obtained in step S3, the pedestrian trajectory is semantically parsed to obtain a semantic trajectory organized and managed in units of local semantics.

[0070] The experimental trajectory is disassembled through the semantic encoding function of the semantic parser to obtain the semantic encoding corresponding to each trajectory, indoor scene elements, the floor where the topological relationship changes, the moving mode adopted by the moving object, the moving object encoding, the key point numbers of the complete trajectory arranged in time, the time when the topological relationship occurs, and other spatio-temporal semantics for subsequent semantic spatio-temporal query and retrieval of the pedestrian trajectory. As Figure 4 shown.

[0071] Select the Yanghu Tianjie Commercial Center in Changsha. After querying and verifying the applicability of the model to the indoor trajectory after compressing the semantic information table of the experimental trajectory integrated in step S4, semantic query results are obtained through the semantic parser and semantic parsing is performed. The semantic spatio-temporal query and retrieval are divided into two types, namely the behavior activities of a single moving object and the behavior activities of multiple moving objects. For the spatio-temporal query and retrieval of the two types, 7 specific indoor query scenarios and 8 query instances based on behavioral topological semantics for the corresponding scenarios are proposed. A query result semantic parsing algorithm is defined to parse and express the semantics corresponding to the query instances. Finally, the query results are displayed and visualized.

[0072] such as Figure 5As shown, the trajectory 4 has the behavioral semantics of entering "Sexy Tea" (POI25) during this time period. If you want to know the common interest preferences of mobile objects for a certain POI indoors, then for the scenario of "mining mobile objects with the same movement behavior in specified indoor scene elements", the query question "During the period from 13:00 on June 15, 2022 to 15:00 on June 15, 2022, which mobile objects have been to 'Sexy Tea' (POI25)?" is proposed to find out which mobile objects are interested in Cha Yan Yue Se.

[0073] As Figure 6 shown, the mobile object 5 has a movement trajectory of entering "Ten O'Clock Bookstore" (POI35) at "2022-06-15 14:48:48". If you want to know the specific time when a mobile object enters a certain POI in an indoor place, then for the scenario of "determining the movement time of different mobile objects in different regions", the query question "When did the mobile object 5 enter the interior of the 'Ten O'Clock Bookstore' (POI35) area?" can be proposed. If a malfunction occurs in indoor public facilities and you want to know which mobile objects are inside, then for the scenario of "determining the movement methods taken by different mobile objects going up and down the stairs", the query question "During the period from 14:00 on June 15, 2022 to 15:30 on June 15, 2022, which mobile objects took the direct elevator downstairs?" can be proposed, and the semantic query results are as Figure 7 shown, it is queried that the mobile objects with semantic encoding I2I1I1 and movement method elevator during this time period are the mobile objects with TrajID 2 and 5. The semantic analysis is as Figure 8 shown, which explains the specific semantic content of the mobile objects 2 and 5 taking the direct elevator downstairs. The query results are visualized as Figures 9 - 12 shown, Figure 9 、 10 The behavioral path of the mobile object 2 moving along the indoor corridor u on the second floor to the direct elevator, taking the direct elevator downstairs to the first floor, and then moving along the indoor corridor f is visualized, Figure 11 、 12 and the behavioral path of the mobile object 5 moving along the indoor corridor u on the second floor to the direct elevator, taking the direct elevator downstairs to the first floor, and then moving along the indoor corridor f is also visualized.

Claims

1. A method for compressing big data of mobile object trajectories for indoor scene semantics, characterized in that, The method includes the following steps: S1. Obtain indoor scene data. The indoor scene includes in-floor scenes and cross-floor scenes. Model important spatial entity elements in the in-floor scenes and cross-floor scenes to obtain topological components of indoor scene elements. S2. Construct an indoor scene local overlapping semantic model by using scene topological tuples established from mobile object trajectory topological tuples and topological components of indoor scene elements to capture the dynamic behavior semantics of trajectories in indoor scene elements. The mobile object trajectory topological tuple consists of a front sub-trajectory, a key point, and a rear sub-trajectory. The indoor scene elements include indoor local corridor elements and indoor area elements. The indoor scene local overlapping semantic model includes a sub-trajectory tuple - indoor local corridor element overlapping semantic model and a sub-trajectory tuple - indoor area element overlapping semantic model constructed by using three topological invariants in the mobile object trajectory topological tuple. S3. Use an encoding method consisting of characters to encode the overlapping semantics between the sub-trajectory tuple and indoor scene elements to obtain effective overlapping semantics of the sub-trajectory for the indoor scene local overlapping semantic model. S4. Reconstruct the trajectory semantics according to the indoor scene local overlapping semantic model. Organize the semantic information in the model in the time dimension in an encoding method, and compress the original trajectory into a complete semantic trajectory. In step S4, the indoor scene local overlapping semantic model compresses the mobile trajectory in the same type of encoding method for indoor local corridor elements and indoor area elements. The general expression form of the indoor scene mobile trajectory is: where Tcode represents the semantic encoding of sub-trajectories with different elements of the indoor scene, "@" is a connection symbol connecting indoor scene elements, "Q" is the unique number of the sub-trajectory tuple corresponding to a certain indoor scene element, " / " is a connection symbol connecting indoor floor numbers, "F" is the unique number representing the floor where the topological component of the key point of the sub-trajectory is located, "*" is a connection symbol connecting the moving modes of the mobile object, and "W" is the moving mode adopted by the mobile object. Use "+" to connect different semantics of the topological tuple where the key point of the sub-trajectory is located, and use "↓" to connect the combined semantics of the floor change of the sub-trajectory between indoor area elements. The reconstruction of the trajectory semantics includes a semantic trajectory reconstruction mode based on indoor local corridor elements and a semantic trajectory reconstruction mode based on indoor area elements, and their trajectory formulas are respectively: where the encoding of the sub-trajectory tuple and the indoor local corridor element is marked in the form of xyz, and x, y, z respectively represent the topological overlapping relationships between the front sub-trajectory, the key point, the rear sub-trajectory and the corridor element. The encoding of the sub-trajectory tuple and the indoor area element is marked in the form of XYZ, and X, Y, Z respectively represent the topological overlapping relationships between the front sub-trajectory, the key point, the rear sub-trajectory and the indoor area element.

2. The method for compressing big data of moving object trajectories for indoor scene semantics according to claim 1, characterized in that, In step S1, perform data preprocessing on the indoor scene data, including format conversion and coordinate conversion.

3. The method for compressing big data of mobile object trajectories oriented to indoor scene semantics according to claim 1, wherein In step S1, for the scene within the floor, the attribute information of the POI area, corridor and public area on each floor is obtained, and for the cross-floor scene, the attribute information of the elevator, escalator and elevator on each floor is obtained to obtain the spatiotemporal information of the indoor scene for semantic trajectory query.

4. The method for compressing big data of moving object trajectories for indoor scene semantics according to claim 1, wherein In step S2, the moving object trajectory is composed of sub-trajectories and key points. The key points are locations where topological relationships change during pedestrian movement. The sub-trajectory that occurs earlier than the time attribute of the key point in the neighborhood of the key point is the front sub-trajectory, and the sub-trajectory that occurs later than the time attribute of the key point is the back sub-trajectory.

5. The method for compressing big data of moving object trajectories oriented to indoor scene semantics according to claim 1, wherein In step S2, the indoor local corridor is composed of multiple corridors in the indoor place, the two ends of each corridor are used as its nodes, the nodes are used as the boundaries of each corridor, the arc segments between the boundaries are used as the interior of the corridor, and the topological space outside the boundaries is used as the exterior of the corridor. The topological components of the indoor local corridor element are composed of the exterior, the boundary and the interior; the topological components of the indoor area element are composed of the front, the back, the exterior and the boundary. The front is the low-floor area of ​​the same scene, the back is the high-floor area of ​​the same scene, and the exterior and boundary are the boundary and exterior of the corresponding areas of the front and back, respectively.

6. The method for compressing big data of moving object trajectories for indoor scene semantics according to claim 1, wherein In step S2, the semantic situation of the overlap between each topological component in the sub-trajectory tuple and the topological component in the indoor local corridor element is calculated as follows: Wherein, the three topological invariants in the sub-trajectory tuple are respectively encoded as the pre-sub-trajectory IT pre , the key-point IT key , and the post-sub-trajectory IT post , L 1 and L 2 respectively represent node 1 and node 2 at both ends of the corridor, and L 0 represents the interior of the corridor except for the nodes; wherein, the default positive moving direction on the corridor is from node 1 to node 2, and the opposite direction is the reverse direction; each value in the matrix represents the overlapping situation between the sub-trajectory topological invariant and the corridor element topological invariant, and its value is empty or 1, -1, 0. Empty means there is an empty set in the two types of topological invariants, 1 and -1 respectively represent a positive intersection and a reverse intersection between the two types of topological invariants, and 0 means there is no intersection between the two types of topological invariants; The semantics of the overlap calculation of each topological component in the sub-trajectory tuple and the topological component in the indoor area element is as follows: wherein, DR d , DR u , DR o and DR - respectively represent the front, i.e., the low floor area, the back, i.e., the high floor area, the boundary, and the external area of the same scene; each value in the matrix represents the overlapping situation between the topological invariant of the sub-trajectory and the topological invariant of the indoor area element, and its value can be empty or 0, 1. Empty means that there is an empty set in the two types of topological invariants, 0 means that there is no intersection between the two types of topological invariants, and 1 means that there is an intersection between the two types of topological invariants.

7. The method for compressing big data of moving object trajectories for indoor scene semantics according to claim 6, characterized in that, In step S3, the encoding method of the sub-trajectory tuple and the indoor local corridor element includes the valid overlapping semantics in the 9 overlapping parts, and the lowercase letters i, b, and e are used to represent the internal overlap of the moving object and the corridor, the boundary overlap of the moving object and the corridor, and the external overlap of the moving object and the corridor respectively; the node numbers b1 and b2 are used to distinguish the movement direction of the sub-trajectory in the corridor, so far i represents the attribute of the intersection of the sub-trajectory and the corridor from the node b1 to b2, and -i represents the attribute of the intersection of the sub-trajectory and the corridor from the node b2 to b1; if there is no null in the two types of topological invariants, the non-null corresponding mark is i, -i, b1, b2, e, otherwise it is marked as null; the encoding of the sub-trajectory tuple and the indoor local corridor element is marked in the form of xyz, and x, y, and z respectively represent the topological overlapping relationship between the front sub-trajectory, the key point, the rear sub-trajectory and the corridor element; The encoding method of the sub-trajectory tuple and the indoor area element includes the valid overlapping semantics in the 16 overlapping parts, and the capital letters I, B, and E are used to represent the internal overlap between the moving object and the indoor area element, the boundary overlap between the moving object and the indoor area element, and the external overlap between the moving object and the indoor area element respectively; the internal numbers I1 and I2 of the area are used to distinguish the sub-trajectories inside the low-floor area and the high-floor area in the same scene. If there is no empty space in the two types of topological invariants, the non-empty space is marked as I1, I2, B, and E respectively, otherwise it is marked as empty; the encoding of the sub-trajectory tuple and the indoor area element is marked in the form of XYZ, and X, Y, and Z respectively represent the topological overlapping relationship between the front sub-trajectory, the key point, the back sub-trajectory and the indoor area element.

Images