Space-time stay point extraction method and system based on multi-source data fusion

Abstract

The invention discloses a space-time stay point extraction method and system based on multi-source data fusion, and aims at carrying out accurate fusion on mobile phone positioning data based on Wi-Fi and GPS, effectively solving a space-time conflict problem and accurately extracting an individual space-time stay point. The method comprises the following steps: acquiring Wi-Fi and GPS positioning data of a target individual and preprocessing the Wi-Fi and GPS positioning data; recording and screening the preprocessed positioning data, and deleting redundant records with the same starting time; speed abnormal data is eliminated, and a positioning data set without speed conflict is generated; performing space-time consistency fusion on the positioning data to generate observation data without space-time conflicts; and a two-stage extraction method is adopted, initial stay points are extracted through spatial distances and time thresholds, clustering is carried out, and final stay points of the individuals are generated. According to the method, the precision of multi-source data fusion and the robustness of stay point extraction are remarkably improved, and an efficient and reliable spatio-temporal data analysis tool is provided for researchers, planners and developers.

Description

Technical Field

[0001] The present invention relates to the technical field of geographic information processing and data fusion, and in particular to a method and system for extracting spatiotemporal stay points based on multi-source data fusion. Background Art

[0002] With the development of mobile communications and positioning technologies, mobile phone positioning data (such as Wi-Fi and GPS data) has been widely used in individual behavior analysis, urban activity pattern analysis, traffic behavior research, and location-based service optimization. Wi-Fi positioning data records the start and end times of connections, making it suitable for analyzing dwelling behavior; GPS positioning data records real-time locations, making it suitable for capturing dynamic trajectories. However, these two types of data often conflict in temporal and spatial dimensions. For example, records with overlapping time but distant locations make it difficult to extract dwelling points.

[0003] Traditional stop point extraction methods often rely on a single data source, failing to fully exploit the complementary nature of multi-source data. Existing fusion methods often rely on extensive manual annotation, which is costly and lacks universal applicability. Furthermore, the lack of a unified standard for handling spatiotemporal conflicts can easily lead to inaccurate extraction results. Therefore, developing an efficient and universal method for spatiotemporal stop point extraction based on multi-source data fusion is crucial for improving data analysis accuracy and application value. Summary of the Invention

[0004] This paper aims to address the spatiotemporal conflict problem in multi-source positioning data fusion, providing a spatiotemporal stop point extraction method and system based on Wi-Fi and GPS data fusion. This method generates accurate individual stop points through data preprocessing, record screening, speed anomaly removal, spatiotemporal consistency fusion, and a two-stage stop point extraction process. It demonstrates strong universality and efficiency.

[0005] To achieve the above object, the present invention is implemented through the following technical solutions:

[0006] A spatiotemporal stay point extraction method based on multi-source data fusion, comprising:

[0007] Data collection and preprocessing: Obtain the target individual's Wi-Fi and GPS positioning data. Wi-Fi data includes a desensitized ID, longitude, latitude, start time, and end time. GPS data includes a desensitized ID, longitude, latitude, and start time. Convert the data to a unified spatial coordinate system (such as WGS-84) and assign the start time and end time to the GPS data as the recording time.

[0008] Record filtering: Delete redundant records with the same start time, retain the longest duration record for Wi-Fi data, and retain a random record for GPS data.

[0009] Speed anomaly elimination: Calculate the speed of each data point. If the speed exceeds the threshold of 500km / h, the point is deleted.

[0010] Spatiotemporal consistency fusion: Centered on the record with the longest duration, based on the spatial distance threshold of 100 meters and the temporal overlap rule, the temporal overlapping data are fused to generate conflict-free observations.

[0011] Two-stage stay point extraction: the first stage extracts preliminary stay points based on a spatial distance of 100 meters and a time threshold of 30 minutes; the second stage aggregates the stay points through bottom-up clustering (distance threshold 150 meters) to generate the final results.

[0012] Preferably, data collection and preprocessing include:

[0013] Use Wi-Fi access points to collect Wi-Fi positioning data;

[0014] Use GPS module to collect high-precision positioning data;

[0015] Use Python or ArcGIS tools to convert the data to a unified coordinate system and standardize the timestamps.

[0016] Preferably, the calculation formula for speed anomaly elimination is:

[0017]

[0018] Among them, v i represents the velocity of the data point, Represents the spatial distance between a data point and the previous data point without time overlap, and Respectively represent the start time of the data point and the end time of the data point;

[0019] Preferably, the spatiotemporal consistency fusion includes:

[0020] Sort the time overlapping data by duration from longest to shortest, and select the longest record as the center;

[0021] If the distance between other data points and the center is less than 100 meters, they are grouped together; otherwise, the retention is determined based on time overlap and speed consistency (speed < 500 km / h);

[0022] Update the start time within the group to the minimum value, the end time to the maximum value, and the position to the medoid.

[0023] Preferably, the two-stage dwell point extraction comprises:

[0024] Phase 1: Data points are represented as (spatial location, start time) and (spatial location, end time). If a group of points lasts for more than 30 minutes within a 100-meter range, it is considered a preliminary stop point.

[0025] The second stage: bottom-up clustering is used to merge the stay points with a distance less than 150 meters, and the distance between clusters is calculated based on medoid.

[0026] To achieve the above-mentioned object, the second aspect of the present invention provides a spatiotemporal stay point extraction system based on multi-source data fusion, comprising:

[0027] Data acquisition module, used to obtain Wi-Fi and GPS positioning data of the target individual and perform pre-processing;

[0028] A screening module is used to screen the positioning data with the same format, delete the records with the same start time, and generate non-redundant positioning data;

[0029] An abnormality elimination module is used to delete speed abnormality data from the non-redundant positioning data set to generate positioning data without speed conflicts;

[0030] A fusion module, configured to perform spatiotemporal consistency fusion on the positioning data set without speed conflict to generate data without spatiotemporal conflict;

[0031] The stay point extraction module is used to extract preliminary stay points from the data without spatiotemporal conflict based on a two-stage extraction method by using spatial distance and time thresholds and cluster them to generate individual final stay points, so as to achieve accurate extraction of spatiotemporal stay points.

[0032] The present invention has at least the following technical effects:

[0033] (1) The present invention effectively solves the spatiotemporal conflict problem between Wi-Fi and GPS data and improves the accuracy of stay point extraction.

[0034] (2) The method of the present invention is simple, relies on a small number of parameters (distance and time thresholds), and is applicable to different scenarios and data sizes.

[0035] (3) The present invention adopts a two-stage extraction method, taking into account the spatial and temporal characteristics of the stay points and enhancing the robustness of the results.

[0036] (4) The present invention is applicable to the fields of individual behavior analysis, urban activity pattern analysis, traffic behavior research, and location service optimization, and has broad application prospects.

[0037] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 : The overall flow chart of the spatiotemporal stop point based on multi-source data fusion in an embodiment of the present invention includes data collection and preprocessing, screening, anomaly elimination, fusion and extraction steps.

[0039] Figure 2 : A flow chart of the spatiotemporal consistency fusion processing of an embodiment of the present invention, showing the grouping and update logic of time-overlapping data.

[0040] Figure 3 : A schematic diagram of the two-stage spatiotemporal stop point extraction process of an embodiment of the present invention, showing the preliminary extraction and clustering process.

[0041] Figure 4 : A structural block diagram of a spatiotemporal stay point extraction system based on multi-source data fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present embodiment is described in detail below. Examples of the embodiment are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention, but are not to be construed as limiting the present invention.

[0043] The following describes in detail the spatiotemporal stay point extraction method and system based on multi-source data fusion of this embodiment with reference to the accompanying drawings.

[0044] Figure 1 FIG is a flow chart of a method for extracting spatiotemporal stay points based on multi-source data fusion according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0045] Step S1: Obtain the Wi-Fi and GPS positioning data of the target individual and perform pre-processing.

[0046] In this embodiment, the Wi-Fi and GPS positioning data of the target individual are collected by Wi-Fi access points and GPS modules. Wi-Fi records the start and end time, longitude, latitude, and desensitized ID of the individual's connection to Wi-Fi, with a time resolution of seconds; GPS data is collected by the built-in GPS module of the device, recording real-time location (longitude, latitude), recording time, and desensitized ID, with a time resolution of minutes. During the collection process, Wi-Fi data and GPS data are stored in JSON format to ensure the structuring of the data.

[0047] The collected Wi-Fi and GPS data were converted to coordinate systems using the Python GeoPandas library. All longitude and latitude data was uniformly converted to the WGS-84 coordinate system to ensure spatial consistency. The Pandas library was also used to standardize the data's timestamps, aligning the time format to UTC. For GPS data, since it only contains the recording time, both the start and end times were set to the recording time to align with the format of the Wi-Fi data, resulting in a unified time field structure (including start time, end time, longitude, and latitude).

[0048] During the data preprocessing phase, the integrity of Wi-Fi and GPS data must be checked. A Python script is used to iterate through all records, removing observations with missing values (e.g., empty latitude and longitude, missing timestamps). Furthermore, the data is grouped according to the desensitized ID, and the Pandas groupby function is used to separate Wi-Fi and GPS data into independent subsets by individual. This ensures that each set of data corresponds to a single individual, facilitating subsequent individualized analysis and stop point extraction. The preprocessed data is stored in Parquet format, leveraging its efficient columnar storage to optimize subsequent data reading and processing.

[0049] Through the above steps, positioning data with consistent format and high integrity are generated, laying the foundation for subsequent record screening, speed anomaly elimination and spatiotemporal consistency fusion, ensuring the stability and accuracy of the entire spatiotemporal stop point extraction process.

[0050] Step S2: Filter the pre-processed data for records with the same start time to obtain non-redundant positioning data.

[0051] In this embodiment, based on the consistent format and high integrity positioning data generated in step S1, records with the same start time are further filtered to eliminate data redundancy and generate a non-redundant positioning dataset. This step adopts different filtering strategies based on the different characteristics of Wi-Fi and GPS data to ensure an efficient filtering process.

[0052] For Wi-Fi data, we used the Python Pandas library to iterate over all records and identify groups of records with the same start time. Using Pandas' groupby function, we grouped the records by anonymized ID and start time. Within each group, we calculated the duration of each record (end time minus start time). We retained the record with the longest duration and deleted all other redundant records to more accurately reflect individual stay behaviors. For GPS data, since it only records real-time location, we adopted a random retention strategy: within groups of records with the same start time, we randomly retained one record and deleted all other duplicate records to avoid data redundancy.

[0053] Step S3: Delete the speed abnormality data to obtain positioning data without speed conflict.

[0054] In this embodiment, based on the non-redundant positioning data in step S2, the positioning data of each individual is sorted in chronological order, and the sort_values function of Pandas is used to sort them in ascending order by desensitized ID and start time to ensure that the data points are processed one by one in time. All data points of each individual are traversed and the speed of each point is calculated step by step. The speed calculation formula is:

[0055]

[0056] Among them, v i represents the velocity of the data point, Represents the spatial distance between a data point and the previous data point without time overlap, and Respectively represent the start time of the data point and the end time of the data point;

[0057] To accurately calculate the distance, we first perform a projection transformation on the longitude and latitude coordinates of the data points. The longitude and latitude data in the WGS-84 coordinate system are converted to a plane coordinate system (such as UTM projection). The Python GeoPandas library is used to implement the projection transformation and generate plane coordinates (x, y). Then, the distance.euclidean function of the SciPy library is used to calculate the Euclidean distance between the two points. The unit is meter. Time difference The speed v is calculated by parsing the timestamp using Python's time package in seconds. i .

[0058] During the traversal process, the speed calculation is based on the previous retained data point. If the speed of the current data point v i If the speed exceeds the threshold of 500 km / h (approximately 138.89 meters per second), the point is deleted and the process continues with the next data point. The speed calculation for the next point is based on the currently retained data point. By calculating and deleting each point point by point, the retained data set is dynamically updated, ultimately generating a positioning data set free of speed conflicts.

[0059] Step S4: Perform spatiotemporal consistency fusion to obtain data without spatiotemporal conflict.

[0060] In this embodiment, based on the velocity conflict-free positioning data generated in step S3, spatiotemporal consistency fusion is further performed to eliminate spatiotemporal conflicts in the data. Observation data segments with temporal overlap are selected from the data. Each temporally overlapping observation data segment is processed as follows. Figure 2The multi-stage processing flow of this step is shown, including step-by-step filtering and grouping of data to ensure data consistency in time and space. The details are as follows:

[0061] First, sort all observations by duration, using the data with the longest duration as the center of the current group. Check to see if the distance between the remaining data points and the center is less than 100 meters. If so, group the points together, mark the center as the center of the group, and end processing for that segment. If not, continue processing for each data point. For any ungrouped data points, loop through all data points except the center and check for any temporal overlap with the retained data points. If so, proceed to the next step.

[0062] For data points with time overlap, check whether there is only one overlapping group: if there is only one group, further determine the distance between the point and the center of the group. If the distance is less than or equal to 100 meters, assign it to the group; if the distance is greater than 100 meters, delete the point; if there are multiple overlapping groups, check whether the distance between the point and the center of all groups is greater than 100 meters. If so, delete the point; otherwise, check whether there is a speed conflict with other retained points (that is, after adding the point, calculate the speed of each retained point; if it is greater than or equal to 500 km / h, it is considered a conflict): if there is a speed conflict, delete the point; otherwise, assign the point to the group with a distance from the center less than or equal to 100 meters.

[0063] For data points without time overlap, check whether the distance from all centers is greater than 100 meters. If so, and the time of the point falls between the times of other retained data points in a group, delete the point. If not, check whether there is a speed conflict with other retained points. If so, delete the point; otherwise, it becomes the center of a new group. If the distance from a center is less than or equal to 100 meters, check whether the number of centers with a distance less than or equal to 100 meters from the point is 1. If so, delete the point; otherwise, assign the point to the group with the corresponding center.

[0064] After the above processing, the start time of each group is updated to the minimum start time of the data points in the group, the end time is updated to the maximum end time of the data points in the group, and the position is updated to the medoid of the data points in the group. Through the above steps, observation data without temporal and spatial conflicts is generated.

[0065] Step S5: A two-stage spatiotemporal stay point extraction method is used to extract preliminary stay points based on spatial distance and time duration thresholds, and final stay points are generated through clustering.

[0066] In this example, based on the observation data generated in step S4 (consistently formatted and free of spatiotemporal conflicts), a two-stage spatiotemporal stay point extraction method is used to further extract individual stay points and generate a final stay point set. This step ensures the accuracy and physical rationality of stay point extraction by combining spatial distance and temporal duration threshold constraints with a clustering method. Figure 3 The overall process diagram of the two-stage spatiotemporal stop point extraction method is shown, and the specific description is as follows:

[0067] In the first stage, preliminary stay points are extracted based on spatial distance and temporal duration thresholds. The observation data for each individual is traversed chronologically. For each data point A, the first point B that satisfies the temporal threshold is found. Starting at point A, subsequent observation points are examined point by point to find the first point B such that the difference between the start time of point B and the start time of point A exceeds the temporal duration threshold (set as 30 minutes). Subsequently, the maximum spatial distance (diameter) of all observation points between point A and point B (including both points A and B) is calculated. The Euclidean distance between all pairs of points is calculated using the Python SciPy library to find the maximum value. If this maximum distance is less than or equal to 100 meters, the observation sequence from point A to point B is considered a candidate for a preliminary stay point. To determine the complete range of this stay point, the traversal continues from point B backward until a point C is encountered such that the maximum spatial distance between all observation points from point A to point C exceeds 100 meters. At this point, the traversal stops, and the point immediately preceding point C is used as the end point for this stay point. Ultimately, the initial stop point is constructed from a sequence of observations from point A to the point immediately preceding point C. Its location is represented by the medoid (geometric median) of all observations in the sequence. The medoid is calculated using the SciPy library, with the point with the smallest maximum distance as the center. The start and end times of the initial stop point are the earliest start and latest end times in the sequence, respectively. This method allows us to explore all possible initial stop points one by one, ensuring that the spatial range and temporal span of each stop point meet the threshold requirements.

[0068] In the second stage, the preliminary stay points are spatially aggregated using clustering to generate the final stay points. A bottom-up hierarchical clustering approach is employed, initially treating each preliminary stay point as a separate cluster. In each iteration, the two closest clusters are found (the distance is calculated based on the Euclidean distance between the medoids of each cluster). If the diameter of the merged cluster does not exceed a specified cluster distance threshold (set to 150 meters), the two clusters are merged into a new cluster, and the medoid of the new cluster is updated to the geometric median of the merged point set. This process is repeated until the distance between all clusters is greater than 150 meters. Ultimately, each cluster represents a final stay point, whose location is represented by the medoids of all preliminary stay points within the cluster, with the start time and end time being the earliest start time and latest end time of all preliminary stay points within the cluster, respectively.

[0069] Figure 4 FIG is a structural block diagram of a spatiotemporal stop point extraction system based on multi-source data fusion according to an embodiment of the present invention. Figure 4 As shown, the spatiotemporal stay point system 100 based on multi-source data fusion includes a data acquisition module 10, a screening module 20, an abnormality elimination module 30, a fusion module 40 and a stay point extraction module 50 connected in sequence to form a linear processing flow.

[0070] In this system, a data acquisition module 10 is used to collect Wi-Fi positioning data and GPS positioning data of a target individual, and perform spatiotemporal data preprocessing to unify the coordinate system and time format to generate positioning data with a consistent format. A screening module 20 is used to record and screen the positioning data with the consistent format, delete records with the same start time, and generate non-redundant positioning data. An anomaly removal module 30 is used to delete speed anomaly data from the non-redundant positioning data set to generate positioning data without speed conflicts. A fusion module 40 is used to perform spatiotemporal consistency fusion on the positioning data set without speed conflicts to generate data without spatiotemporal conflicts. A stay point extraction module 50 is used to extract preliminary stay points from the data without spatiotemporal conflicts based on spatial distance and time thresholds based on a two-stage extraction method, and cluster them to generate the final stay points of the individual, thereby achieving accurate extraction of spatiotemporal stay points.

[0071] In one embodiment of the present invention, when collecting and preprocessing Wi-Fi and GPS positioning data, the data acquisition module 10 is used to: collect the Wi-Fi and GPS positioning data of the target individual through the Wi-Fi access point and the GPS module, the Wi-Fi data records the start and end time, longitude, latitude and desensitized ID of the individual's Wi-Fi connection, and the GPS data records the real-time location (longitude, latitude), recording time and desensitized ID; use Python's GeoPandas library to convert the data into the WGS-84 coordinate system, use the Pandas library to unify the time format into UTC format, and fill in the end time for the GPS data.

[0072] In one embodiment of the present invention, when filtering the positioning data with the same format, the filtering module 20 is configured to: for Wi-Fi data, traverse all records and retain the record with the longest duration among the records with the same start time; for GPS data, randomly retain a record with the same start time to generate non-redundant positioning data.

[0073] In one embodiment of the present invention, when deleting speed anomaly data, the abnormality elimination module 30 is used to: traverse the data points in chronological order, calculate the speed of each data point, and delete the point if the speed exceeds 500 km / h; in the speed calculation, after transforming the coordinate projection into plane coordinates, use the SciPy library to calculate the Euclidean distance to generate positioning data without speed conflicts.

[0074] In one embodiment of the present invention, the fusion module 40 is used to perform spatiotemporal consistency fusion: sort the data by record duration, select the record with the longest duration as the center, group the data based on the 100-meter distance threshold and time overlap rule, update the start time, end time and location of each group, and generate data without spatiotemporal conflicts.

[0075] In one embodiment of the present invention, the stay point extraction module 50 is used to extract stay points based on a two-stage extraction method: in the first stage, the data points are traversed in chronological order, preliminary stay points are extracted based on a spatial distance of 100 meters and a time threshold of 30 minutes, and the medoid position of each stay point is determined; in the second stage, preliminary stay points with a distance of less than 150 meters are merged through hierarchical clustering to generate final stay points.

[0076] It should be noted that the specific implementation of the spatiotemporal stay point extraction system of the embodiment of the present invention can refer to the specific implementation of the spatiotemporal stay point extraction method described above, and will not be repeated here to avoid redundancy.

[0077] In summary, this paper addresses the complexity and limitations of spatiotemporal conflicts in existing positioning data fusion. By comprehensively considering the spatiotemporal characteristics of Wi-Fi and GPS data, this paper provides a spatiotemporal stop point extraction method based on multi-source data fusion. This method is applicable to different scenarios and individual behavior patterns. This method provides a powerful tool for researchers in fields such as individual behavior analysis and urban activity pattern analysis, helping them better understand and analyze individual spatiotemporal stop behavior.

[0078] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0079] Although the present invention has been described in detail through the above preferred embodiments, it should be understood that the above description is not intended to limit the present invention. After reading the above description, various modifications and substitutions of the present invention will become apparent to those skilled in the art. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A spatiotemporal stay point extraction method based on multi-source data fusion, characterized in that: include: Obtain the target individual's Wi-Fi and GPS location data and pre-process them; Filter the pre-processed positioning data and delete redundant records with the same start time; Eliminate speed anomaly data and generate a positioning data set without speed conflicts; Perform spatiotemporal consistency fusion on positioning data to generate observation data without spatiotemporal conflict; A two-stage extraction method was used to extract preliminary stay points based on spatial distance and time thresholds and cluster them to generate the final stay points of individuals.

2. The spatiotemporal stay point extraction method based on multi-source data fusion according to claim 1, characterized in that: The obtaining and pre-processing of the Wi-Fi and GPS positioning data of the target individual includes: Get Wi-Fi location data, including desensitized ID, longitude, latitude, start time, and end time; Get GPS location data, including desensitized ID, longitude, latitude, and start time; Convert Wi-Fi and GPS data into a unified spatial coordinate system. For GPS data, assign the start time and end time as the recording time to generate a unified format for Wi-Fi and GPS data.

3. The spatiotemporal stay point extraction method based on multi-source data fusion according to claim 1, characterized in that: Deleting redundant records with the same start time includes: For Wi-Fi positioning data, the record with the longest duration among those with the same start time is retained; For GPS positioning data, one of the records with the same start time is randomly retained.

4. The method for extracting spatiotemporal stay points based on multi-source data fusion according to claim 1, wherein: The abnormal speed data is removed, including: Calculate the speed of each data point using the formula: Among them, v i represents the velocity of the data point, Represents the spatial distance between a data point and the previous data point without time overlap, and Respectively represent the start time of the data point and the end time of the data point; When the speed exceeds the threshold of 500 km / h, the data point is deleted.

5. The method for extracting spatiotemporal stay points based on multi-source data fusion according to claim 1, characterized in that: The spatiotemporal consistency fusion includes: The record with the longest duration in the time-overlapping data is selected as the center point; Determine whether the spatial distance between other time overlapping data points and the center point is less than 100 meters; If it is less than 100 meters, it will be assigned to the group where the center point is located; if it is greater than 100 meters, it will be deleted or assigned to another group based on time overlap and speed consistency; Update the earliest start time, latest end time and medoid position of the data in the group.

6. The method for extracting spatiotemporal stay points based on multi-source data fusion according to claim 1, wherein: The two-stage extraction method comprises: Phase 1: Based on the spatial distance threshold of 100 meters and the time threshold of 30 minutes, preliminary stay points are extracted, and the spatial location is represented by medoid; The second stage: bottom-up clustering is used to aggregate preliminary stay points with a distance of less than 150 meters to generate final stay points, and the spatial location is represented by medoid.

7. A spatiotemporal stay point extraction system based on multi-source data fusion, characterized in that: include: Data acquisition module, used to obtain Wi-Fi and GPS positioning data of the target individual and perform pre-processing; A screening module is used to screen the positioning data with the same format, delete the records with the same start time, and generate non-redundant positioning data; An abnormality elimination module is used to delete speed abnormality data from the non-redundant positioning data set to generate positioning data without speed conflicts; A fusion module, configured to perform spatiotemporal consistency fusion on the positioning data set without speed conflict to generate data without spatiotemporal conflict; The stay point extraction module is used to extract preliminary stay points from the data without spatiotemporal conflict based on a two-stage extraction method by using spatial distance and time thresholds and cluster them to generate individual final stay points, so as to achieve accurate extraction of spatiotemporal stay points.

Images