Vehicle parking information detection method and device and electronic equipment

Abstract

The embodiment of the invention discloses a vehicle parking information detection method and device and electronic equipment. According to one specific embodiment, the detection method comprises the steps that parking point detection is carried out based on moving track data of a sample vehicle, parking event information of the sample vehicle is generated, and the moving track data are obtained through a satellite positioning system; determining a first type of monitoring record pair matched with the resident event information according to the collection record of each monitoring device; according to the first type of monitoring record pair and the second type of monitoring record pair, sample data is constructed to be used for detecting whether the collection record of the target monitoring record pair represents the existence of vehicle parking information, and the second type of monitoring record pair is the monitoring record pair which is not matched with the parking event information. The implementation mode is related to a data mining technology, and the vehicle parking information can be detected by utilizing monitoring equipment records. Therefore, the number and the type of analyzable vehicles can be increased.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the technical field of data mining, and in particular to a method, a device and an electronic device for detecting vehicle residence information. Background Art

[0002] Resident area inference is often a common problem in trajectory data mining. Researchers can use the resident area of ​​a moving object to understand the semantics of the location and the mobility of the object. Traditional resident area inference algorithms are usually highly dependent on the sampling accuracy of sensor devices, such as GPS (Global Positioning System) trajectories with sampling intervals of minutes and seconds. With the vigorous development of monitoring equipment in cities, the inventors found that monitoring equipment can also be used instead of GPS trajectories to perceive vehicles in the city.

[0003] However, the inventors also found that the traditional resident area detection algorithm is designed for high-sampling rate GPS trajectories. Their core logic is to find subsequences from the original trajectory that satisfy the requirement of staying in a certain area for a specified period of time. However, the sparseness of the monitoring equipment records is often much higher than that of the GPS trajectories, which makes the traditional method ineffective, that is, it cannot be applied to the recorded data of the monitoring equipment.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0005] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0006] Some embodiments of the present disclosure propose a method for detecting vehicle residence information, a device for detecting vehicle residence information, an electronic device, a computer-readable medium, and a computer program product to solve one or more of the technical problems mentioned in the above background technology section.

[0007] In a first aspect, some embodiments of the present disclosure provide a method for detecting vehicle residence information, including: performing residence point detection based on movement trajectory data of a sample vehicle to generate residence event information of the sample vehicle, wherein the movement trajectory data is obtained using a satellite positioning system; determining a first type of monitoring record pair that matches the residence event information based on collection records of each monitoring device, wherein the monitoring record pair is a collection record generated when two monitoring devices successively collect the same vehicle information in a time sequence; constructing sample data based on the first type of monitoring record pair and the second type of monitoring record pair to detect whether the target monitoring record pair represents the existence of vehicle residence information, wherein the second type of monitoring record pair is a monitoring record pair that does not match the residence event information.

[0008] In some embodiments, detecting whether the target monitoring record pair indicates the presence of vehicle residence information includes: using sample data and a travel speed estimation method to detect whether the collected records of the target monitoring record pair indicate the presence of a vehicle residence event.

[0009] In some embodiments, sample data is used to adopt a travel speed estimation method to detect whether the collected records of the target monitoring record pair represent the existence of a vehicle residence event, including: determining the travel speed estimate represented by the target monitoring record pair based on the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair and the time difference in collecting the same vehicle information; determining a speed threshold based on the first travel speed estimate represented by each first-category monitoring record pair and the second travel speed estimate represented by each second-category monitoring record pair in the sample data; and determining whether there is a vehicle residence event based on the comparison result of the travel speed estimate and the speed threshold.

[0010] In some embodiments, detecting whether a target monitoring record pair represents the presence of vehicle residence information further includes: in response to determining that the target monitoring record pair represents the presence of a vehicle residence event, determining parameters of an elliptic curve representing the reachable range of the vehicle based on statistical data of a first type of monitoring record pair in sample data and location information of two target monitoring devices indicated by the target monitoring record pair; determining a set of candidate areas based on the location area covered by the elliptic curve; and selecting at least one candidate area from the set of candidate areas as the residence area corresponding to the vehicle residence event.

[0011] In some embodiments, based on the statistical data of the first type of monitoring record pairs in the sample data and the location information of the two target monitoring devices indicated by the target monitoring record pairs, the parameters of the elliptic curve representing the reachable range of the vehicle are determined, including: performing statistical analysis on the collection records of each first type of monitoring record pair in the sample data to obtain the inverse of the eccentricity of the elliptic curve representing the reachable range of the vehicle; determining the major axis value of the elliptic curve based on the inverse of the eccentricity of the elliptic curve and the straight-line distance between the two target monitoring devices, wherein the straight-line distance between the two target monitoring devices is the minor axis value of the elliptic curve.

[0012] In some embodiments, a candidate area set is determined based on the location area covered by the elliptic curve, including: determining the location area overlapping between the elliptic curve and the pre-divided area grid as a candidate area, and obtaining a candidate area set, wherein the candidate area is a location area whose overlap rate reaches a preset value.

[0013] In some embodiments, at least one candidate area is selected from a candidate area set as a residence area corresponding to a vehicle residence event, including: inputting information of the candidate area set and a target monitoring record pair into a pre-trained residence area selection model, and outputting a residence area prediction result; determining the residence area corresponding to the vehicle residence event according to the residence area prediction result; wherein the residence area selection model is used to predict the residence probability of each candidate area in the candidate area set for the vehicle indicated by the input monitoring record pair, and the residence area selection model is trained using sample data.

[0014] In some embodiments, the resident area selection model includes a candidate area representation module, a spatiotemporal context encoding module and a knowledge fusion module; the candidate area representation module is used to generate a representation vector for each candidate area in the candidate area set; the spatiotemporal context encoding module is used to combine the target monitoring record pairs with other spatiotemporal information and encode the spatiotemporal context vector; the knowledge fusion module is used to use an attention mechanism to fuse the spatiotemporal context vector with the representation vector of each candidate area respectively to obtain the resident probability of each candidate area, and in the order of the resident probability from high to low, the first preset number of candidate areas are used as resident areas and output.

[0015] In some embodiments, generating a representation vector for each candidate area in the candidate area set includes: for each candidate area in the candidate area set, using the TF-IDF statistical method to determine the importance of each set interest point to the candidate area to obtain the representation vector of the candidate area.

[0016] In some embodiments, a residence point detection is performed based on the movement trajectory data of the sample vehicle to generate the residence event information of the sample vehicle, including: preprocessing the movement trajectory data of the sample vehicle to obtain processed trajectory data; performing a residence point detection on the processed trajectory data to generate the residence event information of the sample vehicle.

[0017] In some embodiments, a dwell point detection is performed on the processed trajectory data to generate dwell event information of the sample vehicle, including: traversing each trajectory point in the processed trajectory data in trajectory order, determining whether the distance between the current trajectory point and the previous trajectory point is less than a distance threshold; in response to determining that it is less than the distance threshold, adding the current trajectory point to a candidate trajectory point set; in response to determining that it is greater than the distance threshold, performing a dwell point detection on each candidate trajectory point in the candidate trajectory point set to generate dwell event information of the sample vehicle.

[0018] In some embodiments, a dwell point detection is performed on each candidate trajectory point in the candidate trajectory point set to generate dwell event information of the sample vehicle, including: determining whether the time interval between the first candidate trajectory point and the last candidate trajectory point in the candidate trajectory point set is greater than a time threshold; in response to determining that it is greater than the time threshold, determining each candidate trajectory point in the candidate trajectory point set as a dwell point, and clearing the candidate trajectory point set; extracting the center position, start time and end time of each dwell point as the dwell event information.

[0019] In a second aspect, some embodiments of the present disclosure provide a device for detecting vehicle residence information, including: a residence point detection unit, configured to perform residence point detection based on movement trajectory data of a sample vehicle, and generate residence event information of the sample vehicle, wherein the movement trajectory data is obtained using a satellite positioning system; a monitoring record pair matching unit, configured to determine a first type of monitoring record pair that matches the residence event information based on the collection records of each monitoring device, wherein the monitoring record pair is a collection record generated when two monitoring devices successively collect the same vehicle information in a time sequence; a sample data construction unit, configured to construct sample data based on the first type of monitoring record pair and the second type of monitoring record pair, so as to detect whether the target monitoring record pair represents the existence of vehicle residence information, wherein the second type of monitoring record pair is a monitoring record pair that does not match the residence event information.

[0020] In some embodiments, the detection device further includes a residence information detection unit, which includes a residence event detection subunit configured to use sample data and a travel speed estimation method to detect whether the collected records of the target monitoring record pair indicate the existence of a vehicle residence event.

[0021] In some embodiments, the residence event detection subunit is further configured to determine an estimated travel speed represented by the target monitoring record pair based on the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair and the time difference for collecting the same vehicle information; determine a speed threshold based on the first estimated travel speed represented by each first-category monitoring record pair and the second estimated travel speed represented by each second-category monitoring record pair in the sample data; and determine whether a vehicle residence event exists based on the comparison result of the travel speed estimate and the speed threshold.

[0022] In some embodiments, the residence information detection unit also includes a residence area detection subunit, which is configured to determine, in response to determining that a target monitoring record pair represents the presence of a vehicle residence event, the parameters of an elliptic curve representing the reachable range of the vehicle based on statistical data of a first type of monitoring record pair in sample data and location information of two target monitoring devices indicated by the target monitoring record pair; determine a set of candidate areas based on the location area covered by the elliptic curve; and select at least one candidate area from the set of candidate areas as the residence area corresponding to the vehicle residence event.

[0023] In some embodiments, the resident area detection subunit is further configured to perform statistical analysis on the collection records of each first-category monitoring record pair in the sample data to obtain the inverse of the eccentricity of the elliptic curve representing the reachable range of the vehicle; determine the major axis value of the elliptic curve based on the inverse of the eccentricity of the elliptic curve and the straight-line distance between the two target monitoring devices, wherein the straight-line distance between the two target monitoring devices is the minor axis value of the elliptic curve.

[0024] In some embodiments, the resident area detection subunit is further configured to determine the location area overlapping between the elliptic curve and the pre-divided area grid as a candidate area, and obtain a candidate area set, wherein the candidate area is a location area whose overlap rate reaches a preset value.

[0025] In some embodiments, the resident area detection subunit is further configured to input the information of the candidate area set and the target monitoring record pair into a pre-trained resident area selection model, and output a resident area prediction result; determine the resident area corresponding to the vehicle residence event according to the resident area prediction result; wherein the resident area selection model is used to predict the residence probability of each candidate area in the candidate area set for the vehicle indicated by the input monitoring record pair, and the resident area selection model is trained using sample data.

[0026] In some embodiments, the resident area selection model includes a candidate area representation module, a spatiotemporal context encoding module and a knowledge fusion module; the candidate area representation module is used to generate a representation vector for each candidate area in the candidate area set; the spatiotemporal context encoding module is used to combine the target monitoring record pairs with other spatiotemporal information and encode the spatiotemporal context vector; the knowledge fusion module is used to use an attention mechanism to fuse the spatiotemporal context vector with the representation vector of each candidate area respectively to obtain the resident probability of each candidate area, and in the order of the resident probability from high to low, the first preset number of candidate areas are used as resident areas and output.

[0027] In some embodiments, the candidate region representation module is further configured to determine the importance of each set interest point to the candidate region using the TF-IDF statistical method for each candidate region in the candidate region set to obtain a representation vector for the candidate region.

[0028] In some embodiments, the residence point detection unit is further configured to pre-process the movement trajectory data of the sample vehicle to obtain processed trajectory data; perform residence point detection on the processed trajectory data to generate residence event information of the sample vehicle.

[0029] In some embodiments, the residence point detection unit is further configured to traverse each trajectory point in the trajectory data in trajectory order, determine whether the distance between the current trajectory point and the previous trajectory point is less than a distance threshold; in response to determining that it is less than the distance threshold, add the current trajectory point to the candidate trajectory point set; in response to determining that it is greater than the distance threshold, perform residence point detection on each candidate trajectory point in the candidate trajectory point set to generate residence event information of the sample vehicle.

[0030] In some embodiments, the dwell point detection unit is further configured to determine whether the time interval between the first candidate trajectory point and the last candidate trajectory point in the candidate trajectory point set is greater than a time threshold; in response to determining that it is greater than the time threshold, each candidate trajectory point in the candidate trajectory point set is determined as a dwell point, and the candidate trajectory point set is cleared; and the center position, start time and end time of each dwell point are extracted as dwell event information.

[0031] In a third aspect, some embodiments of the present disclosure provide an electronic device, comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the detection method described in any implementation method in the above-mentioned first aspect.

[0032] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the detection method described in any one of the implementations in the first aspect above is implemented.

[0033] In a fifth aspect, some embodiments of the present disclosure provide a computer program product, including a computer program, which, when executed by a processor, implements the detection method described in any implementation manner in the above-mentioned first aspect.

[0034] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: the vehicle residence information detection method of some embodiments of the present disclosure can use the residence event information contained in the satellite positioning trajectory to help the monitoring device record and infer the residence event and / or residence area. Specifically, the historical movement trajectory data of the same batch of sample vehicles collected can be used to generate residence event information. Then, the monitoring device records corresponding to these vehicles are matched with the residence event information. Thereby, sample data is generated according to the matching results. In this way, vehicle residence detection can be performed using the sample data. That is, whether the target monitoring record is detected to indicate the existence of vehicle residence information. That is, a method for inferring residence information using monitoring device records based on the existing traffic monitoring system is proposed. Thereby, it is beneficial to expand the number and type of vehicles that can be analyzed. It is convenient to timely discover and solve the safety problems in the city, and it helps to reduce the number of safety problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0036] Figure 1 is a flow chart of some embodiments of the vehicle residence information detection method disclosed in the present invention;

[0037] Figure 2A It is a comparison chart of the data collection differences between the satellite positioning system and the monitoring equipment;

[0038] Figure 2B is an architectural diagram of an exemplary system in which some embodiments of the present disclosure may be applied;

[0039] Figure 3A are flow charts of some further embodiments of the vehicle residence information detection method disclosed herein;

[0040] Figure 3B It is a relevant data analysis diagram during the detection of resident events;

[0041] Figure 4A are flow charts of other embodiments of the vehicle residence information detection method disclosed in the present invention;

[0042] Figure 4B and Figure 4C It is a relevant data analysis diagram during the detection process of the resident area;

[0043] Figure 4D is a schematic diagram of the structure of some embodiments of the resident region selection model;

[0044] Figure 5 It is a schematic diagram of the structure of some embodiments of the detection device of vehicle residence information disclosed in the present invention;

[0045] Figure 6 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0046] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0047] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0048] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0049] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0050] With regard to the collection, storage, and use of user personal information (such as user attribute information and user historical operation behaviors) involved in this disclosure, before performing the corresponding operations, the relevant organizations or individuals shall fulfill their obligations, including conducting personal information security impact assessments, fulfilling the obligation to inform the personal information subject, and obtaining the authorization and consent of the personal information subject in advance.

[0051] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0052] Figure 1 The process 100 of some embodiments of the vehicle residence information detection method according to the present disclosure is shown. The detection method comprises the following steps:

[0053] Step 101: Perform residence point detection based on the movement trajectory data of the sample vehicle to generate residence event information of the sample vehicle.

[0054] In some embodiments, the execution entity (such as a server) of the vehicle residence information detection method can obtain the movement trajectory data of the sample vehicle through a wired connection or a wireless connection. The movement trajectory data here is usually collected at a relatively short interval and can more completely reflect the vehicle's movement trajectory. For example, the movement trajectory data can be obtained using a satellite positioning system. The movement trajectory data generally includes the vehicle's identification information (such as a license plate number), the vehicle's location information (such as longitude and latitude, landmark names such as roads or buildings, etc.), the collection time, and so on.

[0055] It is understandable that the satellite positioning system is generally an interrelated collection or device (component) formed with the goal of determining the spatial position. This system can ensure that at any time, at least 4 satellites can be observed at any point on the earth at the same time, so that the satellite can collect the longitude, latitude and altitude of the observation point in order to achieve navigation, positioning, timing and other functions. The above-mentioned satellite positioning system can be at least one of the following: GPS, Galileo satellite navigation system, GLONASS system, BDS (BeiDou NavigationSatellite System, BeiDou Satellite Navigation System), and other satellite positioning systems that may appear in the future.

[0056] In some embodiments, the execution entity can obtain the movement trajectory data of the sample vehicle from the database of the traffic monitoring system. It can also be collected from public resources on the Internet. The sample vehicle here can be any vehicle, such as a private car, a taxi, an online car-hailing vehicle, an ambulance, various engineering transport vehicles, etc. Here, the execution entity can perform a stay point detection on the movement trajectory data of the sample vehicle to generate the stay event information of the sample vehicle. Among them, the stay point usually refers to the location where the vehicle has stayed during operation. The stay event information can be information describing the vehicle's stay situation, such as the start and end time of the stay, the location information of the stay point, etc.

[0057] It is understandable that the execution subject can detect the residence point in a variety of ways. As an example, the execution subject can count the average speed of each sample vehicle passing through the same road section in a similar period of time. In this way, the distance threshold can be obtained based on the average speed and the acquisition frequency (or acquisition interval length) of the satellite positioning system. At this time, for the mobile trajectory data of the same sample vehicle, the execution subject can traverse each trajectory point one by one. And it can be calculated whether the distance between the current trajectory point and the previous trajectory point is less than the distance threshold. If it is less than the distance threshold, the current trajectory point can be determined as a residence point. And the location information (such as longitude and latitude), acquisition start time and acquisition end time of the current trajectory point (or the center between the previous trajectory point) can be used as the residence event information of the sample vehicle.

[0058] Optionally, in order to make the detection result more accurate, the execution subject can also perform residence point detection in combination with the distance threshold and the time threshold. Specifically, first, the execution subject can traverse each track point in the mobile track data in the track order to determine whether the distance between the current track point and the previous track point is less than the distance threshold. If it is determined to be less than the distance threshold, then the current track point can be added to the candidate track point set. If it is determined to be greater than the distance threshold, then the residence point detection can be performed on each candidate track point in the candidate track point set, thereby generating the residence event information of the sample vehicle.

[0059] Secondly, for each candidate trajectory point in the candidate trajectory point set, the execution entity can determine whether the time interval between the first candidate trajectory point and the last candidate trajectory point in the candidate trajectory point set is greater than the time threshold. The first and last candidate trajectory points here are determined according to the trajectory sequence. If it is determined to be greater than the time threshold, each candidate trajectory point in the candidate trajectory point set can be determined as a residence point. At this time, the execution entity can extract the center position, start time (the collection start time of the first candidate trajectory point) and end time (the collection end time of the last candidate trajectory point) of each residence point as the residence event information. In addition, the execution entity can clear the candidate trajectory point set to perform the next residence point detection. The above distance threshold can be set to 50 meters, and the time threshold can be set to 10 minutes.

[0060] In some embodiments, in order to further improve the accuracy of the stationary point detection, before the detection, the execution body can also pre-process the movement trajectory data of the sample vehicle to exclude abnormal positioning data. As an example, the speed and sampling frequency rules can be used to remove abnormal points caused by abnormal positioning of the satellite positioning system, and interpolate between abnormal points. Here, the abnormal points are mainly defined from two perspectives: 1) the instantaneous speed at a certain moment exceeds the threshold; 2) the instantaneous acceleration at a certain moment exceeds the threshold.

[0061] When a trajectory point position suddenly changes due to abnormal positioning of the satellite positioning system, we can determine its starting position through instantaneous acceleration and remove abnormal point clusters from the trajectory in combination with the speed. For example, the previous trajectory point of the section where the instantaneous acceleration begins to change can be determined as the starting position. The next trajectory point of the section until the instantaneous acceleration returns to normal is determined as the end position. In this way, the trajectory is divided into two sections. We perform sequential interpolation by analyzing the velocity and acceleration between the two sections. The interpolation frequency depends on the trajectory accuracy. That is, the interpolation time interval can be determined based on the acquisition frequency. In this way, the interpolation position interval can be determined based on the time interval, velocity and acceleration.

[0062] At this time, the execution subject can use the pre-processed moving trajectory data as the processing trajectory data, and then perform residence point detection on the processing trajectory data to generate residence event information of the sample vehicle.

[0063] Step 102: Determine a first type of monitoring record pair that matches the resident event information based on the collection records of each monitoring device.

[0064] In some embodiments, the execution subject can determine the first type of monitoring record pair that matches the resident event information based on the collection records of each monitoring device. The first type of monitoring record pair here is usually a monitoring record pair that characterizes the presence of a vehicle resident event. Among them, the monitoring record pair (i.e., SCR (surveillance camera record) pair) is usually a collection record generated by two monitoring devices collecting the same vehicle information in time sequence. In other words, the same vehicle passes through the combination of two records generated by two adjacent cameras. The collection records of the monitoring device here can be obtained through the traffic monitoring system. Usually, after obtaining the monitoring video stream, the traffic monitoring system will automatically perform license plate recognition and output the collection records of the monitoring device. A collection record usually includes: license plate number, timestamp, device identification (such as ID) and device location information (such as latitude and longitude). It should be noted that the monitoring device here can be a device that can collect surrounding environment information, especially vehicle-related information, such as a monitoring camera, a speed measuring device, a violation monitoring device, and the like.

[0065] In some embodiments, the execution entity can match the resident event information with the monitoring record pair according to the vehicle identification, start time and end time. Specifically, first, according to the collection records of each monitoring device, the execution entity can filter out the monitoring record pairs collected for the sample vehicle. Then, for the same sample vehicle, when the start timestamps of two consecutive monitoring device records are less than the start time of the resident event information; and the end timestamps of the monitoring device records are greater than the end time of the resident event information. At this time, it can be considered that a pairing is completed. For example, the start time of the previous record in the monitoring record pair is less than the start time of the resident event information, and the end time of the latter record in the monitoring record pair is greater than the end time of the resident event information. At this time, the execution entity can determine the monitoring record pair as a first-class monitoring record pair that matches the resident event information.

[0066] Step 103 : constructing sample data based on the first type of monitoring record pairs and the second type of monitoring record pairs to detect whether the target monitoring record pair represents the presence of vehicle residence information.

[0067] In some embodiments, the execution entity can construct sample data based on the multiple first-category monitoring record pairs and multiple second-category monitoring record pairs obtained in step 102. Among them, the second-category monitoring record pairs are generally monitoring record pairs that do not match the residence event information, that is, monitoring record pairs that characterize the absence of vehicle residence events. For example, the execution entity can use the first-category monitoring record pairs and the corresponding residence event information as positive sample data. And use the second-category monitoring record pairs as negative sample data. In this way, using sample data, it can be used to detect whether the target monitoring record pair characterizes the existence of vehicle residence information. Thereby, the detection of vehicle residence events can be achieved through the monitoring device. The detection of vehicle residence events may include whether a vehicle residence event has occurred, and / or the residence location of the residence event, etc.

[0068] It should be noted that the traditional resident area detection algorithm cannot be applied to monitoring equipment, mainly in the following two aspects:

[0069] First, the uncertainty of the dwelling event: the traditional dwelling area detection algorithm is equivalent to order-based density clustering, which finds the dwelling area by combining the original GPS points. Different from the GPS track, for monitoring equipment, the dwelling area we want to infer is generally independent of the monitoring equipment record. Figure 2AAs shown in the figure on the left, for the GPS trajectory represented by dots, we can use the traditional stay area detection algorithm to identify it. However, for the surveillance equipment records represented by triangles, we cannot use the traditional method to find all available stay events (i.e., Stay Event or Not?) from the surveillance equipment records with a sampling interval of more than 30 minutes. An available stay event is usually set to a stay time of more than 5, 10 or 15 minutes. Not to mention where these stay events occurred (Where to stay? Stay Location).

[0070] The second is the uncertainty of the dwelling area: when a dwelling event occurs, although the monitoring equipment has a wide coverage area, due to budget constraints, most equipment is only deployed at intersections. This means that the dwelling event is often not directly captured by the monitoring equipment. Figure 2A The right figure shows the dwell events between two consecutive monitoring device records and estimates the potential dwell area. Obviously, the farther the distance between the two monitoring devices, the larger the candidate area where the vehicle may reside. In our observation, more than 75% of the recorded potential dwell areas exceed 10 square kilometers.

[0071] It can be seen from the above description that the vehicle residence information detection method in some embodiments of the present disclosure can use the residence event information contained in the satellite positioning trajectory to help the monitoring device record and infer the residence event and / or residence area. Specifically, the historical movement trajectory data collected from the same batch of sample vehicles can be used to generate residence event information. Then, the monitoring device records corresponding to these vehicles are matched with the residence event information. Thereby, sample data is generated according to the matching results. In this way, vehicle residence detection can be performed using the sample data. That is, whether the target monitoring record is detected to indicate the existence of vehicle residence information. That is, a method for inferring residence information using monitoring device records based on an existing traffic monitoring system is proposed. Thereby, it is beneficial to expand the number and types of vehicles that can be analyzed.

[0072] It is understandable that the automatic inference of the residence information based on the monitoring equipment has a wide range of application scenarios, such as discovering illegal activities of special vehicles. For special vehicles, such as chemical transport vehicles and construction waste removal vehicles, the stop event often indicates a potential safety risk. For example, the stop area of ​​a chemical transport vehicle may be a hidden chemical warehouse. The stop area of ​​a construction waste removal vehicle may be related to illegal dumping activities. This helps reduce the number of urban safety issues.

[0073] As an example, Figure 2BAs shown, it shows an architecture diagram of an exemplary system in which some embodiments of the present disclosure can be applied. It is understandable that it is unrealistic to directly use monitoring equipment to infer the resident area without using other information. Therefore, we propose to use the information of the resident area contained in the satellite positioning trajectory to help the monitoring equipment record the inferred resident area. We use the historical movement trajectory data (such as GPS trajectory) collected from the same batch of vehicles and their corresponding monitoring equipment records to construct a data set.

[0074] from Figure 2B It can be seen that the embodiment of the present disclosure designs a framework of a stay inference system SAInf (Stay Area Inference). Among them, the SAInf system can include offline learning (Offline Learning) and online inference (Online Inference) stages. In the offline learning stage, the SAInf system can first construct a data set from the collected monitoring record pairs (SCR Pair) and satellite positioning trajectories (GPS trajectory). Then, the SAInf system learns the relationship between the SCR pairs and their corresponding stay events. As can be seen from the figure, compared with the offline stage, the online stage of the SAInf system does not have the process of constructing a data set, and the SCR pairs to be detected can be directly input. And use the stay event detection module and the stay area identification module trained in the offline stage to online infer whether there is a stay event in the monitoring record pair, and infer where the stay event occurs.

[0075] The entire resident speculation system can include the following three modules:

[0076] First, data preparation module (Data Preparation): In the data preparation stage, it is necessary to pair the stay events in the GPS trajectory data with the SCR pairs to construct a learnable dataset. It includes the extraction of stay events (StayEvent Extraction) and the matching of stay events and SCR pairs (Stay Pair Matching).

[0077] Second, the stay event detection module (Stay Event Detection): Stay event detection uses the travel speed in the SCR pair as a statistic to detect whether there is a stay event in the current SCR pair. It includes the travel speed estimation (Travel Speed ​​Estimation) and the decision boundary estimation (Decision Boundary Estimation)

[0078] Third, the stay area identification module (Stay Area Identification): The stay area identification mainly includes two steps: coarse screening and fine sorting, which correspond to the generation of candidate regions (Candidate Regions Generation) and the selection of stay areas (Stay Area Selection) respectively.

[0079] It should be noted that in the resident inference system of the embodiment of the present disclosure, the processing process of the offline stage includes the online stage. That is, in the offline stage and the online stage, the processing process of the resident event detection module and the stay area identification module is similar. Therefore, the specific processing process can be referred to Figure 3A and Figure 4A The relevant description in will not be repeated here.

[0080] Continue to refer Figure 3A , which shows a process 300 of some other embodiments of the vehicle residence information detection method according to the present disclosure. The detection method may also include the following steps:

[0081] Step 301, determining the estimated value of the travel speed represented by the target monitoring record pair according to the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair and the time difference for collecting the same vehicle information.

[0082] In some embodiments, the execution subject of the vehicle residence information detection method can use sample data and adopt a travel speed estimation method to detect whether the collection record of the target monitoring record pair indicates the existence of a vehicle residence event. The details are as follows:

[0083] It should be noted that in real scenarios, since the number of observation samples between the monitoring device pairs corresponding to the SCR pairs varies greatly, each type of monitoring device pair can be modeled separately. That is, the first type of monitoring record pairs with dwell events and the second type of monitoring record pairs without dwell events are analyzed separately. Based on the idea of ​​statistical modeling, we found that the travel speed of the vehicle can be used to detect whether there is a dwell event in the SCR pair, such as Figure 3B Data analysis diagram shown.

[0084] But a practical problem is that Figure 2B As shown, when we detect the target monitoring record pair online, the input is usually only the information of the SCR pair. This means that the details of the vehicle movement are unknown. So we can use the lower bound of the travel speed as an estimate of the true travel speed. Specifically, we can use the straight-line distance between the two monitoring devices in the SCR pair to replace the true path length. Thus, the following travel speed estimate υ is obtained:

[0085]

[0086] Among them, r i 、r i+1 Respectively represent the two target monitoring devices indicated by the target monitoring record; i .t、r i+1 .t represents the acquisition start time of the two target monitoring devices.

[0087] Step 302: Determine a speed threshold based on a first travel speed estimate represented by each pair of first-category monitoring records and a second travel speed estimate represented by each pair of second-category monitoring records in the sample data.

[0088] In some embodiments, from Figure 3B As can be seen in the figure, the SCR pairs with and without dwell events are visualized and it is found that only this feature (i.e., travel speed estimation) can well distinguish the two types of SCR pairs. However, in specific implementation, we often need a suitable decision boundary for detecting dwell events. Here, the decision boundary is a speed threshold.

[0089] As an example, the execution subject may calculate the average value of each second travel speed estimation value, and use the average value as the speed threshold. Alternatively, the second travel speed estimation values ​​may be clustered and grouped, and the average value or minimum value of the cluster with the largest proportion is determined as the speed threshold.

[0090] Alternatively, if Figure 3B As shown in the figure on the right, we can also use the maximum interval (i.e. diff) of the CDF (Cumulative distribution function) of the two types of SCR pairs. max ) as the speed threshold for decision making (threshold), that is:

[0091]

[0092] Among them, V stay represents the first travel speed estimate; V unstay Represents the second travel speed estimate.

[0093] Step 303: Determine whether a vehicle stay event occurs based on the comparison result between the estimated travel speed and the speed threshold.

[0094] In some embodiments, based on the estimated travel speed value obtained in step 301 and the speed threshold value obtained in step 302, the execution subject can determine whether the target monitoring record pair represents the presence of a vehicle dwelling event according to the comparison result of the two. For example, when the estimated travel speed value between the SCR pair is less than the speed threshold value, it can be considered that a dwelling event has occurred. When the estimated travel speed value between the SCR pair is greater than the speed threshold value, it can be considered that no dwelling event has occurred.

[0095] The method for detecting vehicle dwell information in the disclosed embodiment further enriches and improves the detection process of whether a dwell event occurs between SCR pairs. In view of the uncertainty of dwell events, we analyze the difference in mobility between SCR pairs with dwell events and SCR pairs without dwell events, and use the estimated vehicle travel speed to detect whether a dwell event occurs. Then, the information of the SCR pairs of the monitoring device can be used to detect the dwell event. While ensuring the accuracy of the detection results, it can also reduce the dependence on the satellite positioning system.

[0096] Reference below Figure 4A , which shows a process 400 of another embodiment of the vehicle residence information detection method according to the present disclosure. The detection method may also include the following steps:

[0097] Step 401, in response to determining that a target monitoring record pair represents the presence of a vehicle residence event, determine the parameters of an elliptic curve representing the reachable range of the vehicle based on statistical data of the first type of monitoring record pairs in the sample data and location information of two target monitoring devices indicated by the target monitoring record pair.

[0098] In some embodiments, when it is determined that the target monitoring record pair represents the presence of a vehicle parking event, the execution subject can determine the parameters of the elliptic curve representing the vehicle reachable range based on the statistical data of the first type of monitoring record pairs in the sample data and the location information of the two target monitoring devices indicated by the target monitoring record pair. It can be understood that for the target monitoring record pair, the execution subject can first use the above Figure 3A The detection method shown in the figure performs the resident event detection. When it is determined that there is a resident event, step 401 can be continued. Optionally, the user can also mark the target monitoring record pair with the resident event, and then send the marked target monitoring record pair to the execution subject. At this time, the execution subject can determine that the monitoring record pair has a resident event based on the mark, and then execute process 400 to determine the resident area location.

[0099] In some embodiments, in order to reduce the uncertainty of the resident area, this embodiment can choose to use a combination of coarse screening and fine sorting to design a resident area identification module. In the coarse screening, we hope to determine a candidate area through the spatial distribution of resident events to reduce the difficulty of subsequent resident area selection (i.e., fine sorting of the resident area selection model). For an SCR pair, it describes a moving process that determines the starting point and end point within a specified time range. Therefore, naturally, we propose to use an ellipse to model the reachable range of such a fixed starting point and end point.

[0100] Here, we can scale and translate all SCR pairs containing resident events and their corresponding resident events to achieve spatial normalization. The result is as follows: Figure 4B As shown in the left figure. The dark dots in the figure are the normalized monitoring device (VehicleCamera) positions, and the light dots represent stay events (Stay Point). It can be found that the spatial distribution of stay events conforms to the two-dimensional Gaussian distribution, so an ellipse can be used to model it. We assume that the distance from the starting point to the stay area to the end point of the vehicle is 2a, and the direct distance between the SCR pairs containing the stay event is 2c. Considering the expression of the ellipse, if a suitable distance 2a is found, we can generate an elliptic curve based on the ellipse equation and other parameters. Among them, 2a is usually the major axis of the ellipse; 2c is usually the minor axis of the ellipse.

[0101] In some embodiments, the execution entity can obtain the above 2a by counting each pair of first-category monitoring records in the sample data. For example, the execution entity can count the moving driving distances of the sample vehicles indicated by the first-category monitoring records, and determine the average value of these moving driving distances as the major axis value of the elliptic curve. Alternatively, these moving driving distances can also be clustered and grouped. Thereby, the average value of the cluster with the largest proportion is determined as the major axis value of the elliptic curve. In addition, the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair can be determined as the minor axis value of the elliptic curve.

[0102] Furthermore, in order to improve the accuracy of candidate region determination, in some embodiments, an adaptive ellipse that follows the change of the SCR pair can be obtained. Therefore, using a fixed 2a as the threshold is not enough. Alternatively, the distance rate can be used to generate the ellipse. The expression of the distance rate is generally a / c, which represents the inverse of the eccentricity of the ellipse. Figure 4B The right figure in the middle shows that the statistical characteristics of the distance rate a / c and the distance a as statistics are very similar. Therefore, it is reasonable to use the distance rate to control the size of the ellipse. Based on the previous discussion, the distance rate can be used as a statistic to dynamically estimate the parameters of the elliptic curve.

[0103] As an example, first, the collected records of each pair of first-category monitoring records in the sample data can be statistically analyzed to obtain the reciprocal of the eccentricity of the elliptic curve representing the reachable range of the vehicle. Then, according to the reciprocal of the eccentricity of the elliptic curve and the straight-line distance between the two target monitoring devices, the major axis value of the elliptic curve can be determined, that is, the product of the reciprocal of the eccentricity and the straight-line distance.

[0104] In some embodiments, from Figure 4B As can be seen in , it shows some outliers. Most of them are noises caused by missed detection or false detection by monitoring equipment, which will affect the performance of subsequent fine-tuning. Therefore, we can use the 95% quantile distance rate as the threshold to determine the major axis value of the elliptic curve.

[0105] Step 402: Determine a candidate area set based on the location area covered by the elliptic curve.

[0106] In some embodiments, the execution subject may determine the candidate region set based on the location region covered by the elliptic curve. For example, the execution subject may determine the location region covered by the elliptic curve as the candidate region, thereby obtaining the candidate region set.

[0107] Optionally, the user can also pre-grid the urban area. At this time, the execution subject can determine the location area where the elliptic curve overlaps with the pre-divided regional grid as a candidate area, and obtain a set of candidate areas. Among them, the candidate area can be a location area where the overlap rate reaches a preset value. For example, the execution subject can determine the area where the elliptic curve and the grid completely overlap as a candidate area. For another example, the execution subject can determine the area where the overlap rate between the two reaches 75%, or the area where at least three corners of the grid overlap with the elliptic curve as a candidate area.

[0108] Step 403: Select at least one candidate area from the candidate area set as a residence area corresponding to the vehicle residence event.

[0109] In some embodiments, the execution subject may select at least one candidate area from the candidate area set obtained in step 402 as the residence area corresponding to the above vehicle residence event. As an example, the execution subject may count the number of residence events occurring in each candidate area according to the residence event information of each sample vehicle in the sample data. Thus, the first few candidate areas may be determined as residence areas in descending order of the number.

[0110] From the above description, we can see that the candidate region generation module generates a candidate region set for each SCR pair by estimating the threshold. In the residence region selection module, our task is usually to select several regions from the candidate region set. However, due to the complexity of the residence pattern, three challenges arise. The first challenge is how to extract the relationship between regions and residence events. Secondly, considering various spatiotemporal correlations, encoding the spatiotemporal context of the monitoring record pairs of the residence query is another challenge. Finally, how to effectively fuse this information and calculate the stay probability of each region from the candidate region set is another challenge.

[0111] In some optional implementations, this embodiment designs a Stay Net model (i.e., a resident area selection model). The execution subject can use sample data to train the model to obtain a resident area selection model. Among them, the resident area selection model can be used to predict the resident probability of each candidate area in the candidate area set for the vehicle indicated by the input monitoring record pair. The use of model prediction is conducive to improving the accuracy of resident area determination. At this time, the information of the candidate area set and the target monitoring record pair can be input into the pre-trained resident area selection model to output the resident area prediction result. Thus, according to the resident area prediction result, the resident area corresponding to the vehicle stay event can be determined. For example, the model outputs the resident probability of each candidate area. At this time, the execution subject can select at least one candidate area as the resident area in the order of resident probability from large to small.

[0112] It should be noted that the resident area selection model of this embodiment can adopt a commonly used machine learning network structure. As an example, the resident area selection model can adopt a multi-task learning framework. At this time, the model may include an input layer, a multi-gated hybrid expert layer and an output layer. Among them, the multi-gated hybrid expert layer is usually composed of multiple networks called experts. Each expert is generally composed of several layers of fully connected layers, and the number of hidden layer units can be set. In addition, the number of tasks in the model can be set according to the number of target monitoring record pairs. A gate network can be set for each task to output the probability of each expert being selected. The number of experts can be set here according to the number of candidate areas.

[0113] It is understandable that since the candidate regions corresponding to each target monitoring record may not be the same, they are often analyzed one by one. In other words, the number of tasks is usually 1. In this case, the tower layer can be omitted. The output of the multi-gated hybrid expert layer is the input of the output layer. The output layer can select several regions as resident regions and output them.

[0114] In some embodiments, in order to overcome the above challenges, this embodiment may construct a residence area selection model based on the following three viewpoints.

[0115] First, the residence event is affected by the spatiotemporal context. Figure 4C The left figure shows the heat map of vehicle residence at different POIs (Point of Interest) and time periods. The POIs here can be set according to actual conditions, such as the left vertical coordinate in the heat map (Industrial Zone to Shopping). Obviously, vehicles tend to stay at similar types of locations during similar periods. For example, hazardous chemical vehicles need to transport chemicals, and they often stay near companies, stores, medical and other facilities. Taxis tend to visit urban hot spots frequently. Therefore, the spatiotemporal context information of the monitoring record pairs of the resident query can be captured by time period, visited monitoring equipment, weather, etc.

[0116] Second, the dwelling behavior is intentional and semantically relevant to the region. We model regions with POIs to further capture the relationship between dwelling events and regions.

[0117] at last, Figure 4C The figure on the right shows the average cosine similarity within each candidate region set. It indicates that the discrimination of the region sets is very low, because the average cosine similarity of more than half of the region sets exceeds 0.5. Therefore, it is very challenging to select regions containing resident events from the candidate region set. Here, this embodiment can use the powerful information interaction capability in the transformer model to help extract the differences between regions. In general, the model calculates the resident probability of each candidate region in the candidate region set and returns the regions with the highest resident probability.

[0118] Here, if Figure 4DAs shown, the resident region selection model may include a candidate region representation module (CandidateRegion Representation), a spatio-temporal context encoding module (Spatio-temporal Context Encoding) and a knowledge fusion module (Knowledge Fusion). Among them, the candidate region representation module can be used to generate representation vectors of each candidate region in the candidate region set; the spatio-temporal context encoding module can be used to combine the target monitoring record pair with other spatio-temporal information to encode the spatio-temporal context vector; the knowledge fusion module can be used to adopt an attention mechanism to fuse the spatio-temporal context vector with the representation vector of each candidate region respectively, to obtain the resident probability of each candidate region, and in the order of resident probability from high to low, the first preset number of candidate regions are used as resident regions and output.

[0119] Specifically, in the candidate region representation module, for each candidate region in the candidate region set, the TF-IDF statistical method can be used to determine the importance of each set interest point to the candidate region, so as to obtain the representation vector of the candidate region. Among them, TF-IDF is usually the abbreviation of term frequency-inverse document frequency, which is a commonly used weighting technique for information retrieval and data mining. TF is usually the term frequency, and IDF is usually the inverse text frequency index. In other words, the TF-IDF vector of the POI can be used to uniformly model all regions in the candidate region set and output the variable representation of each region. The processing process is as follows:

[0120]

[0121]

[0122] Among them, X r represents the candidate region set; E r Representation of each candidate region.

[0123] In the spatiotemporal context encoding module, the spatiotemporal information of each SCR pair can be extracted and embedded, and concatenated to generate a hidden representation that indicates the spatiotemporal context. The process is as follows:

[0124] e w =Embed(x w )

[0125] e b =Embed(x b )

[0126] e st =FFN([x s ;e w;e b ;x d ]);

[0127] Among them, x w Indicates weather characteristics; x b 、x d Indicates time characteristics; x s Represents the spatial characteristics of the monitoring equipment.

[0128] In the knowledge fusion module, each region representation can be fused with the spatiotemporal background representation (spatial-temporal context encoding) in an adaptive manner and further enhanced to achieve ideal prediction results.

[0129] E st =MultiHead(e st , E r , e st )+E r

[0130]

[0131] From the above description, we can see that the processing flow of the Stay Net model is as follows: Stay Net takes the target surveillance record pairs (SCR pairs) and their corresponding candidate regions as input. First, the candidate region representation module takes the candidate region set as input and generates a representation of each region. Then, the spatiotemporal context encoding module encodes the time (Time of Day, Workday or Not), location (Camera location Feature) and other information (such as weather Weather) of the SCR pair into a spatiotemporal context vector. Finally, the spatiotemporal context vector is used as a query and key, and the representation of each region is sent to the knowledge fusion module as a value. For each region, the knowledge fusion module can fuse this information and return the residence probability based on the attention mechanism. Here, the BCE Loss (Binary Cross Entropy Loss) binary cross entropy loss function can be used for model training.

[0132] The vehicle residence information detection method of the disclosed embodiment further enriches and improves the residence area detection process. In view of the uncertainty of the residence area, this embodiment designs a residence area inference algorithm that combines coarse screening and fine sorting. In the coarse screening process, the spatial distribution of residence events relative to the monitoring device pair can be modeled under the spatial normalization scale, and the residence candidate area set corresponding to the SCR pair can be generated. In the fine sorting process, the deep neural network Stay Net can be used to calculate the residence probability of each candidate area. Finally, a preset number of areas before the residence probability is output.

[0133] It can be seen that the disclosed embodiment proposes a method for inferring the residence area using monitoring equipment records. We construct a data set using the historical GPS data of the same batch of vehicles collected and their corresponding monitoring equipment records. Among them, SCR pairs are used as input data, and the residence area is extracted from the GPS trajectory as the ground truth. The model captures the residence pattern in different time and space contexts by modeling the relationship between the residence area truth and the SCR pairs. Based on this idea, this embodiment designs a framework of a residence inference system SAInf. SAInf consists of two stages: the system first detects whether a residence event occurs; then it infers where the residence event occurs. The two-stage design can effectively alleviate the uncertainty of residence events and residence areas. This method can automatically provide candidate residence areas and their patrol priorities, effectively improving the duty efficiency of law enforcement departments.

[0134] Further references Figure 5 , as a response to the above Figures 1 to 4D In order to realize the detection method shown in the figure, the present disclosure provides some embodiments of a detection device for vehicle parking information. These detection device embodiments are similar to Figures 1 to 4D The detection method embodiments shown correspond to the above. The detection device for vehicle residence information can be specifically applied to various electronic devices.

[0135] like Figure 5As shown, some embodiments of the vehicle residence information detection device 500 may include: a residence point detection unit 501, configured to perform residence point detection based on the movement trajectory data of the sample vehicle to generate the residence event information of the sample vehicle, wherein the movement trajectory data is obtained by using a satellite positioning system; a monitoring record pair matching unit 502, configured to determine a first type of monitoring record pair that matches the residence event information based on the collection records of each monitoring device, wherein the monitoring record pair is a collection record generated by two monitoring devices collecting the same vehicle information in a time sequence; a sample data construction unit 503, configured to construct sample data based on the first type of monitoring record pair and the second type of monitoring record pair, so as to detect whether the target monitoring record pair represents the existence of vehicle residence information, wherein the second type of monitoring record pair is a monitoring record pair that does not match the residence event information.

[0136] In some embodiments, the detection device 500 may also include a residence information detection unit (not shown in the figure), which may include a residence event detection subunit (not shown in the figure), configured to use sample data and a travel speed estimation method to detect whether the collected records of the target monitoring record pair indicate the existence of a vehicle residence event.

[0137] In some embodiments, the residence event detection subunit is further configured to determine an estimated travel speed represented by the target monitoring record pair based on the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair and the time difference for collecting the same vehicle information; determine a speed threshold based on the first estimated travel speed represented by each first-category monitoring record pair and the second estimated travel speed represented by each second-category monitoring record pair in the sample data; and determine whether a vehicle residence event exists based on the comparison result of the travel speed estimate and the speed threshold.

[0138] In some embodiments, the residence information detection unit also includes a residence area detection subunit, which is configured to determine, in response to determining that a target monitoring record pair represents the presence of a vehicle residence event, the parameters of an elliptic curve representing the reachable range of the vehicle based on statistical data of a first type of monitoring record pair in sample data and location information of two target monitoring devices indicated by the target monitoring record pair; determine a set of candidate areas based on the location area covered by the elliptic curve; and select at least one candidate area from the set of candidate areas as the residence area corresponding to the vehicle residence event.

[0139] In some embodiments, the resident area detection subunit is further configured to perform statistical analysis on the collection records of each first-category monitoring record pair in the sample data to obtain the inverse of the eccentricity of the elliptic curve representing the reachable range of the vehicle; determine the major axis value of the elliptic curve based on the inverse of the eccentricity of the elliptic curve and the straight-line distance between the two target monitoring devices, wherein the straight-line distance between the two target monitoring devices is the minor axis value of the elliptic curve.

[0140] In some embodiments, the resident area detection subunit is further configured to determine the location area overlapping between the elliptic curve and the pre-divided area grid as a candidate area, and obtain a candidate area set, wherein the candidate area is a location area whose overlap rate reaches a preset value.

[0141] In some embodiments, the resident area detection subunit is further configured to input the information of the candidate area set and the target monitoring record pair into a pre-trained resident area selection model, and output a resident area prediction result; determine the resident area corresponding to the vehicle residence event according to the resident area prediction result; wherein the resident area selection model is used to predict the residence probability of each candidate area in the candidate area set for the vehicle indicated by the input monitoring record pair, and the resident area selection model is trained using sample data.

[0142] In some embodiments, the resident area selection model may include a candidate area representation module, a spatiotemporal context encoding module and a knowledge fusion module; the candidate area representation module is used to generate a representation vector for each candidate area in the candidate area set; the spatiotemporal context encoding module is used to combine the target monitoring record pairs with other spatiotemporal information and encode the spatiotemporal context vector; the knowledge fusion module is used to use an attention mechanism to fuse the spatiotemporal context vector with the representation vector of each candidate area respectively to obtain the resident probability of each candidate area, and in order of the resident probability from high to low, the first preset number of candidate areas are used as resident areas and output.

[0143] In some embodiments, the candidate region representation module may be further configured to determine the importance of each set interest point to the candidate region using the TF-IDF statistical method for each candidate region in the candidate region set, so as to obtain a representation vector for the candidate region.

[0144] In some embodiments, the stay point detection unit 501 may be further configured to pre-process the moving trajectory data of the sample vehicle to obtain processed trajectory data; perform stay point detection on the processed trajectory data to generate stay event information of the sample vehicle.

[0145] In some embodiments, the residence point detection unit 501 can be further configured to traverse each trajectory point in the trajectory data in trajectory order, determine whether the distance between the current trajectory point and the previous trajectory point is less than a distance threshold; in response to determining that it is less than the distance threshold, add the current trajectory point to the candidate trajectory point set; in response to determining that it is greater than the distance threshold, perform residence point detection on each candidate trajectory point in the candidate trajectory point set to generate residence event information of the sample vehicle.

[0146] In some embodiments, the dwell point detection unit 501 can be further configured to determine whether the time interval between the first candidate trajectory point and the last candidate trajectory point in the candidate trajectory point set is greater than a time threshold; in response to determining that it is greater than the time threshold, determine each candidate trajectory point in the candidate trajectory point set as a dwell point, and clear the candidate trajectory point set; extract the center position, start time and end time of each dwell point as dwell event information.

[0147] It is understandable that the various units recorded in the vehicle stationary information detection device 500 are similar to those in the reference Figures 1 to 4D Therefore, the operations, features and beneficial effects described above for the method are also applicable to the detection device 500 and the units included therein, and will not be described in detail here.

[0148] Reference below Figure 6 , which shows a structural schematic diagram of an electronic device 600 suitable for implementing some embodiments of the present disclosure. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0149] like Figure 6 As shown, the electronic device 600 may include a processing device 601 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0150] Typically, the following devices may be connected to the I / O interface 605: input devices 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 607 including, for example, a speaker, a vibrator, etc.; storage devices 608 including, for example, a disk, a hard disk, etc.; and communication devices 609. The communication devices 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Figure 6 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 6 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0151] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0152] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0153] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0154] The computer-readable medium may be included in the electronic device; or it may exist independently without being installed in the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: performs residence point detection based on the movement trajectory data of the sample vehicle and generates residence event information of the sample vehicle, wherein the movement trajectory data is obtained using a satellite positioning system; determines a first type of monitoring record pair that matches the residence event information based on the collection records of each monitoring device, wherein the monitoring record pair is a collection record generated by two monitoring devices collecting the same vehicle information in a time sequence; constructs sample data based on the first type of monitoring record pair and the second type of monitoring record pair to detect whether the target monitoring record pair represents the existence of vehicle residence information, wherein the second type of monitoring record pair is a monitoring record pair that does not match the residence event information.

[0155] In addition, computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0156] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0157] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The units described may also be provided in a processor, for example, may be described as: a processor including a residence point detection unit, a monitoring record pair matching unit, and a sample data construction unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the residence point detection unit may also be described as a "unit for performing residence point detection based on the movement trajectory data of the sample vehicle".

[0158] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0159] Some embodiments of the present disclosure further provide a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned vehicle residence information detection methods.

[0160] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A method for detecting vehicle residence information, comprising: Performing residence point detection based on movement trajectory data of a sample vehicle to generate residence event information of the sample vehicle, wherein the movement trajectory data is obtained using a satellite positioning system; Determine, according to the collection records of each monitoring device, a first type of monitoring record pair that matches the resident event information, wherein the monitoring record pair is a collection record generated when two monitoring devices successively collect the same vehicle information in a time sequence; Based on the first and second types of monitoring record pairs, sample data is constructed to detect whether the target monitoring record pair represents the presence of vehicle residence information, wherein the second type of monitoring record pair is a monitoring record pair that does not match the residence event information.

2. The detection method according to claim 1, wherein: The detection target monitoring record indicates whether there is vehicle parking information, including: The sample data is used to adopt a travel speed estimation method to detect whether the collection records of the target monitoring record pair represent the existence of a vehicle residence event.

3. The detection method according to claim 2, wherein: The method of using the sample data and adopting a travel speed estimation method to detect whether the collected records of the target monitoring record pair represent the presence of a vehicle residence event includes: Determine an estimated value of the travel speed represented by the target monitoring record pair based on the straight-line distance between the two target monitoring devices indicated by the target monitoring record pair and the time difference between collecting the same vehicle information; Determine a speed threshold based on a first travel speed estimate represented by each pair of first-category monitoring records and a second travel speed estimate represented by each pair of second-category monitoring records in the sample data; Based on the comparison result of the travel speed estimate and the speed threshold, it is determined whether a vehicle stay event exists.

4. The detection method according to claim 1, wherein: The detection target monitoring record indicates whether there is vehicle parking information, and further includes: In response to determining that the target monitoring record pair represents the presence of a vehicle residence event, determining parameters of an elliptic curve representing a reachable range of the vehicle based on statistical data of the first type of monitoring record pairs in the sample data and location information of two target monitoring devices indicated by the target monitoring record pair; Determine a candidate area set based on the location area covered by the elliptic curve; At least one candidate area is selected from the candidate area set to serve as a stay area corresponding to the vehicle stay event.

5. The detection method according to claim 4, wherein: The determining, based on the statistical data of the first type of monitoring record pairs in the sample data and the position information of the two target monitoring devices indicated by the target monitoring record pairs, the parameters of the elliptic curve representing the reachable range of the vehicle include: Performing statistical analysis on the collected records of each pair of first-category monitoring records in the sample data to obtain the inverse of the eccentricity of the elliptic curve representing the reachable range of the vehicle; The major axis value of the elliptic curve is determined according to the inverse of the eccentricity of the elliptic curve and the straight-line distance between the two target monitoring devices, wherein the straight-line distance between the two target monitoring devices is the minor axis value of the elliptic curve.

6. The detection method according to claim 4, wherein: The step of determining a candidate area set based on the location area covered by the elliptic curve comprises: The location area where the elliptic curve overlaps with the pre-divided area grid is determined as a candidate area to obtain a candidate area set, wherein the candidate area is a location area where the overlap rate reaches a preset value.

7. The detection method according to claim 4, wherein: The selecting at least one candidate area from the candidate area set as the stationary area corresponding to the vehicle stationary event includes: Input the information of the candidate area set and the target monitoring record pair into a pre-trained resident area selection model, and output a resident area prediction result; Determining a retention area corresponding to the vehicle retention event according to the retention area prediction result; The resident area selection model is used to predict the resident probability of each candidate area in the candidate area set for the vehicle indicated by the input monitoring record, and the resident area selection model is trained using the sample data.

8. The detection method according to claim 7, wherein: The resident region selection model includes a candidate region representation module, a spatiotemporal context encoding module and a knowledge fusion module; The candidate region representation module is used to generate a representation vector for each candidate region in the candidate region set; The spatiotemporal context encoding module is used to combine the target monitoring record pair with other spatiotemporal information to encode and obtain a spatiotemporal context vector; The knowledge fusion module is used to adopt the attention mechanism to fuse the spatiotemporal context vector with the representation vector of each candidate area respectively, to obtain the residence probability of each candidate area, and to output the first preset number of candidate areas as residence areas in order of residence probability from high to low.

9. The detection method according to claim 8, wherein: The generating a representation vector of each candidate region in the candidate region set includes: For each candidate region in the candidate region set, the importance of each set interest point to the candidate region is determined by using the TF-IDF statistical method to obtain a representation vector of the candidate region.

10. The detection method according to any one of claims 1 to 9, wherein: The performing of the stationary point detection based on the movement trajectory data of the sample vehicle to generate the stationary event information of the sample vehicle includes: Preprocessing the moving trajectory data of the sample vehicle to obtain processed trajectory data; Performing residence point detection on the processed trajectory data to generate residence event information of the sample vehicle.

11. The detection method according to claim 10, wherein: The performing residence point detection on the processed trajectory data to generate residence event information of the sample vehicle includes: Traversing each track point in the processed track data in track order, and determining whether the distance between the current track point and the previous track point is less than a distance threshold; In response to determining that the distance is less than a threshold, adding the current trajectory point to a set of candidate trajectory points; In response to determining that the distance is greater than a distance threshold, performing a stay point detection on each candidate trajectory point in the candidate trajectory point set to generate stay event information of the sample vehicle.

12. The detection method according to claim 11, wherein: The performing a stay point detection on each candidate trajectory point in the candidate trajectory point set to generate stay event information of the sample vehicle includes: Determine whether a time interval between a first candidate trajectory point and a last candidate trajectory point in the candidate trajectory point set is greater than a time threshold; In response to determining that the time is greater than the time threshold, determining each candidate trajectory point in the candidate trajectory point set as a dwell point, and clearing the candidate trajectory point set; The center position, start time and end time of each of the dwell points are extracted as the dwell event information.

13. A device for detecting vehicle parking information, comprising: A stay point detection unit is configured to perform stay point detection based on movement trajectory data of a sample vehicle and generate stay event information of the sample vehicle, wherein the movement trajectory data is obtained using a satellite positioning system; The monitoring record pair matching unit is configured to determine a first type of monitoring record pair that matches the resident event information according to the collection records of each monitoring device, wherein the monitoring record pair is a collection record generated when two monitoring devices successively collect the same vehicle information in a time sequence; The sample data construction unit is configured to construct sample data based on the first-category monitoring record pair and the second-category monitoring record pair for detecting whether the target monitoring record pair represents the existence of vehicle residence information, wherein the second-category monitoring record pair is a monitoring record pair that does not match the residence event information.

14. An electronic device comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the detection method as described in any one of claims 1-12.

15. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the detection method according to any one of claims 1 to 12 is implemented.

16. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the detection method according to any one of claims 1 to 12.

Images