Off-road area positioning method, electronic device and vehicle

Abstract

The application provides a method for positioning an off-road area, an electronic device and a vehicle. The method comprises the following steps: screening a second driving track representing off-road behavior from a plurality of first driving tracks of vehicles; identifying a first off-road point from the second driving track, and determining a candidate area based on the first off-road point; screening a seed area from the candidate area; and clustering the seed area multiple times to determine the off-road area. In the present scheme, the second driving track representing off-road behavior is screened, and then the candidate area is determined by using the second driving track. The seed area is screened from the candidate area, and the seed area is clustered multiple times to determine the off-road area. The off-road area is identified by using the off-road point without relying on a preset geometric template, so that a large number of non-off-road areas are not incorrectly introduced, and the identification effect and positioning accuracy of the off-road area are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a positioning method, electronic device, and vehicle for off-road areas. Background Technology

[0002] With the rise of off-road culture, off-road enthusiasts have an increasing demand for information on legal off-road venues, which necessitates the identification and location of off-road areas.

[0003] Currently, geometric fitting is commonly used to identify off-road areas, which uses the smallest bounding rectangle or circle of the vehicle trajectory points as the boundary of the off-road area. However, the shape and height of off-road areas are irregular, and using the smallest bounding rectangle or circle for fitting can introduce a large number of non-off-road areas or miss the real off-road areas, resulting in low recognition and positioning accuracy of off-road areas. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method for locating off-road areas, an electronic device, and a vehicle to solve the problems of low recognition effect and low positioning accuracy in existing methods for identifying off-road areas.

[0005] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] The first aspect of this invention discloses a method for locating an off-road area, the method comprising:

[0007] Select the second driving trajectory representing off-road behavior from the first driving trajectories of multiple vehicles;

[0008] Identify the first off-road point from the second driving trajectory, and determine the candidate area based on the first off-road point;

[0009] Seed regions are selected from the candidate regions;

[0010] The seed region is clustered multiple times to determine the off-road region.

[0011] Preferably, a second driving trajectory representing off-road behavior is selected from the first driving trajectories of multiple vehicles, including:

[0012] Using a pre-built whitelist of vehicles with off-road characteristics, off-road vehicles can be identified from multiple vehicles;

[0013] The motion characteristics of the first driving trajectory of the off-road vehicle are verified.

[0014] A second driving trajectory representing off-road behavior is selected, wherein the second driving trajectory is the first driving trajectory of the off-road vehicle verified by the motion characteristics.

[0015] Preferably, identifying a first off-road point from the second driving trajectory and determining a candidate region based on the first off-road point includes:

[0016] Extract the center point of each trajectory segment in the second driving trajectory to obtain the first off-road point;

[0017] Perform coordinate projection transformation on the first off-road point;

[0018] A preset region is constructed centered on the first off-road point where the coordinate projection transformation has been completed, and a candidate region is obtained.

[0019] Preferably, selecting a seed region from the candidate regions includes:

[0020] Query the environmental attribute parameters of the candidate region;

[0021] Seed regions are selected from the candidate regions, where the seed regions are the candidate regions whose environmental attribute parameters meet the preset filtering conditions.

[0022] Preferably, the seed region is clustered multiple times to determine the off-road region, including:

[0023] The seed region is subjected to loose clustering to obtain the first clustering result;

[0024] The first clustering result is subjected to multi-layer progressive filtering to obtain discrete second off-road points;

[0025] The second off-road location is precisely clustered to determine the off-road area.

[0026] Preferably, querying the environmental attribute parameters of the candidate region includes:

[0027] Call the map interface to query the preset element data within the candidate area;

[0028] The preset element data is parsed and classified statistically analyzed to obtain the environmental attribute parameters of the candidate region.

[0029] Preferably, an off-road vehicle is identified from multiple vehicles using a pre-built whitelist of vehicles with off-road characteristics, including:

[0030] Obtain multiple off-road determination conditions built into a pre-built whitelist of vehicles with off-road characteristics;

[0031] Identify off-road vehicles from among multiple vehicles that meet any one of the aforementioned off-road determination criteria.

[0032] Preferably, the center points of each trajectory segment in the second driving trajectory are extracted to obtain the first off-road point, including:

[0033] The second driving trajectory is sliced ​​and sampled to obtain the trajectory center point of a single continuous trajectory segment;

[0034] The center point of a single continuous trajectory segment is determined as the first off-road point. The first off-road point includes at least latitude and longitude coordinates, sampling time, trajectory duration, terrain elevation, and distance from paved road.

[0035] A second aspect of this invention discloses a positioning system for off-road areas, the system comprising:

[0036] The first filtering unit is used to filter out the second driving trajectory representing off-road behavior from the first driving trajectories of multiple vehicles.

[0037] The processing unit is used to identify a first off-road point from the second driving trajectory and determine a candidate region based on the first off-road point;

[0038] The second filtering unit is used to filter out seed regions from the candidate regions;

[0039] Clustering units are used to perform multiple clustering operations on the seed region to determine the off-road region.

[0040] Preferably, the first screening unit includes:

[0041] The identification module is used to identify off-road vehicles from multiple vehicles by utilizing a pre-built whitelist of vehicles with off-road characteristics.

[0042] The verification module is used to verify the motion characteristics of the first driving trajectory of the off-road vehicle.

[0043] The first filtering module is used to filter out the second driving trajectory that represents the occurrence of off-road behavior. The second driving trajectory is the first driving trajectory of the off-road vehicle that has passed the motion feature verification.

[0044] A third aspect of the present invention discloses an electronic device, comprising: a processor and a memory, wherein the processor and the memory are connected via a communication bus; wherein the processor is used to call and execute a program stored in the memory; and the memory is used to store the program, the program being used to implement the off-road area positioning method disclosed in the first aspect of the present invention.

[0045] A fourth aspect of the present invention discloses a vehicle, the vehicle including the electronic equipment disclosed in the third aspect of the present invention.

[0046] Based on the above embodiments of the present invention, a method, electronic device, and vehicle for locating off-road areas are provided. The method involves: filtering a second driving trajectory representing off-road behavior from a first driving trajectory of multiple vehicles; identifying a first off-road point from the second driving trajectory and determining a candidate area based on the first off-road point; filtering a seed area from the candidate area; and performing multiple clustering operations on the seed area to determine the off-road area. In this solution, a second driving trajectory representing off-road behavior is filtered out, and then a candidate area is determined using the second driving trajectory. A seed area is filtered out from the candidate area, and multiple clustering operations are performed on the seed area to determine the off-road area. This method does not rely on a preset geometric template; it identifies the off-road area based on the off-road point, avoiding the erroneous introduction of a large number of non-off-road areas and improving the identification effect and positioning accuracy of the off-road area. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0048] Figure 1 A flowchart illustrating a method for locating an off-road area according to an embodiment of the present invention;

[0049] Figure 2 A flowchart for filtering out the second driving trajectory provided in an embodiment of the present invention;

[0050] Figure 3 A flowchart for determining candidate regions provided in an embodiment of the present invention;

[0051] Figure 4 A flowchart for determining the off-road area provided in an embodiment of the present invention;

[0052] Figure 5 This is a structural block diagram of a positioning system for off-road areas provided in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0055] Off-road driving refers to driving on unpaved terrain such as deserts, mud, snow, and water crossings, away from paved roads. With the rise of off-road culture, off-road enthusiasts have an increasing demand for information on legal off-road areas. Map manufacturers, car companies, and self-driving tour platforms urgently need automated, low-cost, and high-precision methods for identifying and locating off-road areas.

[0056] Research has found that common methods for identifying off-road areas mainly include manual annotation, single-dimensional automatic identification, and geometric fitting.

[0057] Manual annotation method: POI (Point of Interest) information uploaded by off-road clubs and off-road enthusiast communities is manually edited and verified before being entered into the electronic map.

[0058] Single-dimensional automatic identification method: Use the POI database to search for keywords (such as keywords such as "off-road" and "desert") to roughly determine off-road behavior, or simply judge off-road behavior based on whether the vehicle trajectory leaves the road network of the map software.

[0059] Geometric fitting method: The minimum bounding rectangle or bounding circle of the vehicle trajectory points is used as the boundary of the off-road area.

[0060] However, firstly, relying on manual input results in extremely incomplete POIs for off-road areas, failing to reflect the true activity level of the areas, and update cycles are typically measured in years. Secondly, relying solely on POI keyword searches will miss a large number of unnamed off-road terrains, and relying solely on "whether it leaves the road network" will misclassify temporary parking lots, construction sites, and village paths as off-road areas. Thirdly, the shapes and heights of off-road areas are highly irregular (distributed along valleys, ridges, and riverbanks), and fitting them with the minimum bounding rectangle or circle will introduce a large number of non-off-road areas (false positives) or cut off real off-road areas (false negatives).

[0061] To address the aforementioned problems, this invention proposes a method, electronic device, and vehicle for locating off-road areas. The method involves filtering out a second driving trajectory representing off-road behavior, and then using this trajectory to determine candidate areas. Seed areas are then selected from the candidate areas, and these seed areas are clustered multiple times to determine the actual off-road areas. Identifying off-road areas using off-road locations avoids incorrectly introducing a large number of non-off-road areas, improving the identification effect and positioning accuracy of off-road areas, and eliminating the need for manual intervention in off-road area identification.

[0062] See Figure 1 The flowchart illustrates a method for locating an off-road area according to an embodiment of the present invention, the method comprising:

[0063] Step S101: Select the second driving trajectory representing off-road behavior from the first driving trajectories of multiple vehicles.

[0064] In the specific implementation step S101, vehicle IoT trajectory data is collected second by second, which includes the first driving trajectory of multiple vehicles.

[0065] It should be noted that the vehicle IoT trajectory data is the positioning trajectory data transmitted back by compliant vehicles at a frequency of seconds. It has the characteristics of high data density, wide coverage and strong real-time performance, and can truly reflect the actual driving path and spatial position changes of the vehicle.

[0066] For example, a rectangular area of ​​latitude and longitude can be defined by the coordinate system boundary of a province, and second-by-second trajectory data of vehicles within this rectangular area can be collected. In other words, by setting a geographical rectangular area, the spatial domain for data collection can be initially defined, reducing the data size required for subsequent computation and processing, thereby improving processing efficiency.

[0067] From the first driving trajectories of multiple vehicles collected, the second driving trajectory that represents off-road behavior is selected, and the second driving trajectory is the valid off-road trajectory.

[0068] For example, taking province A as an example, the rectangular area of ​​latitude and longitude of the province is collected (97.35°E ≤ λ ≤ 108.55°E, 26.05°N ≤ λ). The location trajectories of all compliant vehicles within the ≤ 34.32°N range were continuously tracked for 30 days, totaling approximately 12 million primary driving trajectories. Subsequently, approximately 85,000 vehicles with off-road characteristics (i.e., off-road vehicles mentioned later) were selected using a vehicle whitelist, and the primary driving trajectories of these off-road vehicles were verified for motion characteristics, ultimately resulting in approximately 620,000 secondary driving trajectories.

[0069] Step S102: Identify the first off-road point from the second driving trajectory, and determine the candidate area based on the first off-road point.

[0070] In the specific implementation step S102, a first off-road point is identified from the filtered second driving trajectory as a candidate point. The first off-road point represents the core location where a certain vehicle actually engages in off-road behavior during a certain period. Candidate areas are then determined based on the first off-road point.

[0071] The candidate region represents a set of discrete points where a vehicle actually engages in a specific off-road behavior. It only includes points that meet the characteristics of off-road behavior and objectively records the location where the off-road behavior actually occurs.

[0072] Step S103: Select seed regions from the candidate regions.

[0073] In the specific implementation step S103, the environmental attribute parameters of the candidate area are queried. Specifically, the map interface is called to query the preset feature data (such as location and building map feature data) within the candidate area. The preset feature data is then parsed and classified statistically to obtain the environmental attribute parameters of the candidate area. The map interface can be an open API (Application Programming Interface) provided by various electronic maps, capable of returning detailed information on geographic features such as POIs, buildings, and roads within the candidate area based on spatial location and search radius.

[0074] For example, a unified map interface can be used to perform batch location and building map searches, querying all location and building map feature data within a candidate area at once. After parsing and classifying these feature data, the environmental attribute parameters of the candidate area can be obtained.

[0075] In some embodiments, the environmental attribute parameters of the candidate region are structured parameters, which can be represented as: Qregion(b)={npoi,nbuilding,has_paved_road}.

[0076] Where Qregion(b) represents the environmental attribute parameters of candidate region b.

[0077] The npoi represents the number of artificial points of interest, which can be retrieved using general POI tags. It counts all commercial, public service, and other artificial locations within the candidate area. These locations include artificial POIs such as restaurants, hotels, supermarkets, gas stations, service areas, village locations, and public facilities, and are used to determine whether there are traces of human activity in the area.

[0078] nbuilding represents the number of man-made buildings. It can be searched by building tags to identify and count all man-made structures. Man-made structures include residential houses, factories, construction sites, temporary housing, supporting facilities, and all other man-made buildings. It is used to accurately distinguish between purely natural areas and man-made construction areas.

[0079] `has_paved_road` is a boolean identifier for paved roads. It can be determined by combining road labels and surface attributes to search for the existence of paved roads within the candidate area. Roads with the highway attribute of motorway, trunk, primary, secondary, tertiary, or residential, and the surface attribute of asphalt, concrete, or paved are uniformly identified as paved roads. If a paved road exists, the paved road identifier is marked as true (paved road identifier is 1); if no paved road exists, the paved road identifier is marked as false (paved road identifier is 0).

[0080] Among them, highway represents the road grade attribute, motorway represents the expressway, trunk represents the trunk road, primary represents the main road, secondary represents the secondary road, tertiary represents the tertiary road, residential represents the residential road, surface represents the road surface material attribute, asphalt represents the asphalt pavement, concrete represents the concrete pavement, and paved represents the paved pavement.

[0081] In practical applications, a data denoising mechanism is used when querying the environmental attribute parameters of candidate areas. This mechanism filters out duplicate features, invalid or expired features, and features with labeling errors, ensuring that the three core fields of npoi, nbuilding, and has_paved_road are accurate and effective. This provides reliable underlying data support for subsequent seed area selection and pseudo-offroad scene removal.

[0082] The data noise reduction mechanism specifically includes: deduplicating and merging identical features recorded repeatedly within the same geographic coordinate range; removing invalid features marked as abandoned, demolished, or relocated from the map data; and correcting or deleting erroneous features with significant deviations in latitude and longitude labels (such as deviating from the road by more than a reasonable threshold).

[0083] After retrieving the environmental attribute parameters of the candidate regions, seed regions are selected based on these parameters. Seed regions are candidate regions whose environmental attribute parameters meet preset filtering criteria. The selected seed regions can be considered as pristine natural wilderness areas.

[0084] For example, the preset filtering criteria are: npoi=0, nbuilding=0, and has_paved_road=0. Candidate regions with "npoi=0, nbuilding=0, and has_paved_road=0" are selected as seed regions. This filtering criterion simultaneously excludes three typical human intervention elements: traces of human activity, artificial buildings, and paved roads, ensuring that the seed regions possess original natural attributes.

[0085] By selecting seed areas from candidate areas, areas with artificial points of interest, artificial buildings, and paved roads can be completely eliminated, leaving only pure natural wilderness areas without artificial facilities (i.e., seed areas), ensuring that candidate areas possess the basic attributes of off-road scenarios.

[0086] Step S104: Perform multiple clustering operations on the seed region to determine the off-road region.

[0087] In the specific implementation step S104, after the seed region is obtained, the seed region is clustered multiple times to determine the off-road region.

[0088] For example, DBSCAN loose clustering and DBSCAN precise clustering are performed on the seed region to determine the off-road region.

[0089] In this embodiment of the invention, a second driving trajectory representing off-road behavior is selected, and then candidate regions are determined using the second driving trajectory. Seed regions are selected from the candidate regions, and multiple clustering operations are performed on the seed regions to determine the off-road regions. This method does not rely on preset geometric templates but identifies off-road regions based on off-road points, avoiding the erroneous introduction of a large number of non-off-road regions and improving the identification effect and positioning accuracy of off-road regions.

[0090] Regarding the above embodiments of the present invention Figure 1 The second driving trajectory involved in step S101, see [link / reference]. Figure 2 The flowchart for filtering out the second driving trajectory provided by an embodiment of the present invention is shown. Figure 2 Includes the following steps:

[0091] Step S201: Identify off-road vehicles from multiple vehicles using a pre-built whitelist of vehicles with off-road characteristics.

[0092] It should be noted that a whitelist of vehicles with off-road characteristics is constructed based on the vehicle's VIN code, original factory model parameters, vehicle configuration attributes, and historical driving tags. This whitelist contains multiple off-road determination criteria; a vehicle that meets any one of these criteria can be identified as an off-road vehicle. The vehicle's VIN code serves as a unique identifier and can be precisely associated with the vehicle's original factory parameters and configuration information, thereby ensuring the accuracy of vehicle attribute determination.

[0093] In the specific implementation step S201, multiple off-road determination conditions are obtained from a pre-built whitelist of vehicles with off-road characteristics. Vehicles that meet any one of the off-road determination conditions are identified from the multiple vehicles to determine off-road vehicles. Specifically, a vehicle that "meets any one of the off-road determination conditions" is considered an off-road vehicle.

[0094] For example, the vehicle whitelist has the following three off-road determination criteria.

[0095] Off-road condition 1: Off-road SUVs, off-road pickups, and professional off-road vehicles equipped with a factory-installed four-wheel drive system and a non-load-bearing body.

[0096] It should be noted that a non-load-bearing body refers to a body structure with an independent frame (steel beam). The body and frame are flexibly connected, allowing it to withstand significant torsional loads, a crucial structural feature of professional off-road vehicles. A four-wheel drive system refers to the vehicle's ability to transmit power to all four wheels, significantly improving its off-road capability and traction.

[0097] Off-road condition 2: Vehicles with professional off-road modification attributes, including differential locks, off-road tires, vehicle lift, and anti-collision modifications.

[0098] It's important to note that a differential lock is a device used to lock the differential, achieving a rigid connection between the two wheels. In complex terrain, it effectively prevents wheel spin and improves off-road capability. Off-road tires refer to specialized tires with deep tread patterns and large block tread blocks, providing stronger grip and puncture resistance. Vehicle lift refers to increasing the vehicle's ground clearance by replacing the suspension system or installing a lift kit to enhance its off-road capability.

[0099] Off-road determination criterion 3: Vehicles that have a long history of continuous driving on unpaved and complex terrain and have been marked as off-road vehicles. This off-road determination criterion 3 is based on the long-term statistical characteristics of the vehicle's historical trajectory data and can cover some original non-off-road models but special vehicles actually used in off-road scenarios.

[0100] For a given vehicle, if the vehicle meets any one of the above off-road determination conditions 1, 2, and 3, then the vehicle can be determined to be an off-road vehicle.

[0101] By identifying off-road vehicles through step S201, driving trajectories of ordinary family cars, urban SUVs, and commercial vehicles that do not have off-road attributes can be eliminated, thus preventing invalid trajectories from interfering with the "determining off-road areas" from the data source.

[0102] Step S202: Verify the motion characteristics of the first driving trajectory of the off-road vehicle.

[0103] It should be noted that after identifying off-road vehicles, not all of their initial travel trajectories can be used to determine the off-road area. In the specific implementation step S202, the initial travel trajectories of the off-road vehicles are subjected to multi-dimensional motion feature verification, thereby filtering out the second travel trajectories that represent off-road behavior from the initial travel trajectories.

[0104] It should be noted that motion feature verification is a dynamic screening mechanism based on trajectory kinematic parameters. By analyzing dimensions such as vehicle speed changes, steering characteristics, spatial position relationships, and terrain undulations during the driving process, it can accurately identify whether the vehicle is actually engaged in off-road driving, rather than simply relying on static judgment based on vehicle type.

[0105] Step S203: Select the second driving trajectory that represents the off-road behavior. The second driving trajectory is the first driving trajectory of the off-road vehicle that has passed the motion feature verification.

[0106] It should be noted that the motion feature verification sets multiple verification conditions. Only when all verification conditions are met simultaneously can the motion feature verification be considered successful. This multi-condition joint judgment method can effectively eliminate false judgments based on only a single condition, ensuring the rigor of trajectory selection.

[0107] In the specific implementation step S203, after performing multi-dimensional motion feature verification on the first driving trajectory of the off-road vehicle, a second driving trajectory representing the off-road behavior is selected. The second driving trajectory is the first driving trajectory of the off-road vehicle that has passed the motion feature verification. The second driving trajectory can be understood as a segment of the actual off-road trajectory.

[0108] For example, the motion feature verification is set with the following four verification conditions.

[0109] Verification condition 1 (verification of driving speed): The speed is maintained at a stable low speed range of 10-40km / h for more than 5 seconds, which is consistent with the characteristics of low-speed driving in complex off-road terrain. Verification condition 1 adopts dual constraints of duration (5 seconds) and speed range (10-40km / h), which can effectively avoid misjudgment caused by instantaneous speed fluctuations. At the same time, the speed range of 10-40km / h covers the typical driving speed in most off-road scenarios.

[0110] Verification Condition 2 (Verification of Steering Characteristics): The steering angle fluctuates significantly, with no prolonged periods of constant-speed straight-line cruising, consistent with the characteristics of off-road driving involving bumps and obstacle avoidance. The steering angle refers to the rate of change of the vehicle's heading angle. During off-road driving, due to terrain undulations and obstacle distribution, vehicles need to frequently steer to avoid obstacles or choose the optimal route, resulting in a high-frequency, large-amplitude fluctuation in the steering angle sequence. In contrast, paved road driving typically features gentler steering angle changes and long straight-line cruising. By analyzing the fluctuation amplitude and frequency of the steering angle, off-road driving and conventional road driving can be effectively distinguished.

[0111] Verification condition 3 (road distance verification): The vehicle continuously moves away from the paved road at a distance of ≥50 meters to exclude pseudo-off-road behaviors such as temporary pull-to-the-side, temporary lane changes, and roadside parking; This verification condition 3 uses paved road network data to calculate the vertical distance from the trajectory point to the nearest paved road. The 50-meter distance threshold can effectively exclude temporary activities at the road edge, while ensuring that the trajectory point truly enters the unpaved area.

[0112] Verification Condition 4 (Terrain Feature Verification): The driving area has an elevation difference ≥ 30 meters, or is located in non-flat and complex terrain such as the Gobi Desert, desert, mountains, or riverbanks. The elevation difference refers to the difference between the highest and lowest elevations within a trajectory segment. A 30-meter elevation difference threshold can distinguish between off-road scenarios with varying terrain and flat roads. Special landforms such as the Gobi Desert, desert, mountains, and riverbanks can be identified through matching with remote sensing data or map terrain labels.

[0113] For the first driving trajectory of an off-road vehicle, if the first driving trajectory meets all four verification conditions mentioned above, it can be determined that the first driving trajectory has passed the motion feature verification, and thus the first driving trajectory can be determined as the second driving trajectory.

[0114] The above embodiments of the present invention Figure 2 This document explains how to select a second driving trajectory. After identifying off-road vehicles using a vehicle whitelist, it abandons simple location judgment and retains only the first driving trajectory of the off-road vehicle that has passed motion feature verification as the second driving trajectory. This second driving trajectory is then used to determine the off-road area, improving the identification effect and positioning accuracy of the off-road area. Specifically, through a dual screening mechanism of vehicle whitelist and motion feature verification, the original second-by-second trajectory data from the Internet of Things is compressed, greatly reducing the data scale of subsequent spatial calculations. Simultaneously, the serial architecture of dual verification ensures that misjudgments at each stage are intercepted in a timely manner, avoiding the cumulative error amplification effect of false off-road trajectories on subsequent point identification and area aggregation. This ensures the accuracy and efficiency of the entire off-road area identification process from the data processing source, maximizing the elimination of invalid interference data without overlooking genuine off-road trajectories.

[0115] Regarding the above embodiments of the present invention Figure 1 For the candidate regions involved in step S102, see [link / reference]. Figure 3 The flowchart illustrating the determination of candidate regions provided in an embodiment of the present invention is shown. Figure 3 Includes the following steps:

[0116] Step S301: Extract the center point of each trajectory segment in the second driving trajectory to obtain the first off-road point.

[0117] In the specific implementation step S301, the second driving trajectory (equivalent to an effective off-road trajectory) is sliced ​​and sampled to obtain the trajectory center point of a single continuous trajectory segment. The trajectory center point of the single continuous trajectory segment is determined as the first off-road point (the point of a single off-road action).

[0118] It should be noted that when extracting the first off-road location, the center point of the trajectory is used instead of the entire trajectory for subsequent processing. This can effectively compress the amount of data and reduce computational complexity. At the same time, the center point, as the spatial centroid of the trajectory segment, can represent the core location where the off-road activity occurred and has strong spatial representativeness.

[0119] The first off-road location includes at least the following basic attributes: latitude and longitude coordinates, sampling time, corresponding vehicle VIN, trajectory duration, terrain elevation, and distance from paved road.

[0120] For example, for a second driving trajectory that lasts 180 seconds and contains 180 second-by-second positioning points, three consecutive trajectory segments are identified based on the fluctuation boundaries of speed and steering angle. The durations of these three consecutive trajectory segments are 45 seconds, 62 seconds, and 73 seconds, respectively. The center point of these three consecutive trajectory segments is taken as three independent first off-road points. Each first off-road point is accompanied by basic attributes such as sampling time, corresponding vehicle VIN, terrain elevation, and distance from the nearest paved road.

[0121] Step S302: Perform coordinate projection transformation on the first off-road point.

[0122] In the specific implementation step S302, after the first off-road point is selected, in order to adapt to spatial distance calculation and clustering operation, and to eliminate the angle unit calculation error of latitude and longitude spherical coordinates, it is necessary to perform coordinate projection transformation on the first off-road point, and uniformly convert WGS-84 latitude and longitude coordinates to EPSG:3857 Web Mercator plane rectangular coordinates, so as to realize the metric standardization calculation of the full spatial data.

[0123] It should be noted that WGS-84 (World Geodetic System 1984) is the standard geographic coordinate system used by the GPS Global Positioning System, and its coordinates are expressed in latitude and longitude (in degrees). EPSG:3857 Web Mercator projection is a normal conformal cylindrical projection widely used in web maps, mapping spherical coordinates to plane rectangular coordinates in meters, facilitating Euclidean distance calculations and spatial cluster analysis.

[0124] Specifically, the coordinate projection transformation of the first off-road point can be performed using formulas (1) and (2).

[0125] (1);

[0126] (2);

[0127] In formulas (1) and (2), R is the Earth's equatorial radius, which is fixed at R=6378137m and is the standard reference parameter of the EPSG:3857 coordinate system. Let be the longitude of the i-th first off-road point in the WGS-84 coordinate system, in degrees. Let be the latitude of the i-th first off-road point in the WGS-84 coordinate system, with coordinates in degrees. The coordinates are the X-axis rectangular coordinates of the Web Mercator plane after coordinate projection transformation, in meters. The coordinates are the Y-axis rectangular coordinates of the Web Mercator plane after coordinate projection transformation, in meters. π is the mathematical constant pi.

[0128] By performing coordinate projection transformation on the first off-road point using the above formulas (1) and (2), all the first off-road points after coordinate projection transformation are uniformly converted to metric plane coordinates, which can avoid the accuracy deviation of latitude and longitude spherical angle calculation and ensure the accuracy of subsequent clustering operations, distance filtering, and regional area calculation, and the results are accurate.

[0129] Step S303: Construct a preset region centered on the first off-road point where the coordinate projection transformation has been completed, and obtain the candidate region.

[0130] In the specific implementation step S303, a preset area is constructed with the first off-road point where the coordinate projection transformation is completed as the center, thereby obtaining the corresponding candidate area.

[0131] For example, a circular retrieval buffer zone with a fixed radius is constructed centered on the first off-road point where the coordinate projection transformation has been completed, thereby obtaining candidate areas. The radius of the buffer zone can be set according to the actual application scenario and data accuracy, for example, it can be set to 100 meters to 500 meters, to ensure that the scope of the buffer zone can cover the actual off-road activity area corresponding to the point, while avoiding excessive expansion that introduces irrelevant features.

[0132] The candidate regions obtained through the above method are without manual annotation or preset region boundaries. They only objectively record all points that actually occur and conform to off-road characteristics, thus providing raw sample data for subsequent calculations to determine off-road regions. These data are temporary data that are dynamically updated in real time and are continuously enriched and refreshed with the continuous access of vehicle IoT data, reflecting the latest distribution trend of off-road behavior.

[0133] The above embodiments of the present invention Figure 3 This section explains how to determine candidate regions. Firstly, when extracting the first off-road checkpoint, the data size is significantly compressed by condensing each trajectory segment into a single center point. This also avoids the interference of spatial autocorrelation between adjacent points within the same trajectory on the clustering results, ensuring that each point in the candidate region comes from independent off-road behavior records. Consequently, the confidence assessment based on the number of points has more reliable statistical significance.

[0134] Secondly, during the coordinate projection transformation of the first off-road point, all points are uniformly transformed to Web Mercator plane rectangular coordinates, so that all subsequent distance thresholds (such as 500-meter neighborhood radius, 5-meter road distance, and 2000-meter cluster radius) are executed based on a unified metric unit, which not only ensures the consistency of calculation accuracy but also improves processing efficiency.

[0135] Thirdly, when constructing candidate regions, a circular buffer zone with a fixed radius is used as the center of the first off-road point after coordinate transformation. The generation of this candidate region depends entirely on the actual location of the off-road behavior, without introducing any artificially preset geometric boundaries or empirical offsets. Compared to the traditional minimum bounding rectangle method or bounding circle method, which forcibly includes a large number of areas where no off-road behavior has occurred as boundaries, the candidate regions constructed by this scheme naturally coincide with the actual location of the off-road behavior. This avoids boundary offsets and area distortions caused by the mismatch between the geometric template and the actual region shape, providing more accurate input data for subsequent region selection and clustering.

[0136] Regarding the above embodiments of the present invention Figure 1 For the off-road area involved in step S104, see [link / reference]. Figure 4 The flowchart for determining the off-road area provided in an embodiment of the present invention is shown. Figure 4 Includes the following steps:

[0137] Step S401: Perform loose clustering on the seed region to obtain the first clustering result.

[0138] It should be noted that DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a classic spatial clustering algorithm that can automatically identify clusters of arbitrary shapes based on the spatial density distribution of points and effectively label isolated noise points.

[0139] Loose clustering refers to a DBSCAN clustering configuration that uses a small neighborhood radius (e.g., ε1=500 meters) and a very low minimum point threshold (e.g., MinPts1=1).

[0140] In the specific implementation step S401, the first round of DBSCAN loose clustering (first clustering) is performed on the center point of the selected seed region to obtain the first clustering result (also known as the candidate group), which contains multiple seed regions.

[0141] In practical applications, the parameters for DBSCAN loose clustering can be set to "ε1=500 meters, MinPts1=1", or the clustering parameters can be set according to the actual situation. ε1 represents the neighborhood radius. A smaller neighborhood radius of 500 meters helps to identify small clusters with relatively discrete spatial distribution. MinPts1 represents the minimum number of points threshold. Setting it to 1 means that a single point can form an initial cluster, which has the maximum inclusiveness.

[0142] When performing DBSCAN loose clustering, a loose clustering strategy with small neighborhoods and no minimum point limit is adopted to maximize the inclusion of remote and sparse niche off-road locations, avoiding the omission of small and unpopular off-road areas and achieving full coverage of potential off-road areas. The core advantage of this strategy is that the primary goal in the first round of clustering is to "not miss anything," including all spatial clusters that may contain off-road activity in the subsequent processing.

[0143] Step S402: Perform multi-layer progressive filtering on the first clustering results to obtain discrete second off-road points.

[0144] In the specific implementation step S402, the first clustering result obtained from the loose clustering is subjected to multi-layer progressive filtering to obtain purified and discrete second off-road points. The multi-layer progressive filtering can employ a layer-by-layer convergence strategy, with each layer targeting a specific type of pseudo-off-road scenario for targeted removal. The filtering intensity gradually increases from coarse to fine, ensuring that effective off-road points are retained to the greatest extent possible during the filtering process.

[0145] In some embodiments, multi-layer progressive filtering includes any and multiple of the following filtering methods: "POI + building + paved road strength filtering", "group size threshold filtering", "road distance check", "IQR outlier removal", "5 km POI check", "off-road vehicle proportion filtering", "trajectory continuity filtering", "terrain compliance filtering", and "secondary clustering noise pre-filtering".

[0146] The following sections will provide a detailed explanation of each of these filtering methods.

[0147] 1. POI + Building + Paved Road Strength Filtering: This layer of filtering is used to eliminate pseudo-off-road areas with traces of artificial construction, based on the above embodiments of the present invention. Figure 1 The environmental attribute parameters obtained in step S103 are used to perform a full-domain validation of the regional attributes corresponding to each first clustering result. This filtering layer is the first line of defense in the spatial environment dimension, excluding all regions with obvious traces of human activity from a geographic semantic perspective.

[0148] The environmental attribute parameter can be represented as: Qregion(b)={npoi,nbuilding, has_paved_road}.

[0149] The core filtering formula for "POI + Building + Paved Road Strength Filtering" is: "npoi=0 and nbuilding=0 and has_paved_road=0".

[0150] Only those that meet the above core screening and judgment formula can enter the next layer of filtering. If any artificial POI, artificial building, or paved road exists, the corresponding first cluster result will be removed. This ensures that the remaining area is a pure natural wilderness scene without human intervention, avoiding interference from pseudo-off-road areas such as villages, factories, and scenic spots from the bottom layer.

[0151] 2. Group Size Threshold Filtering: This layer of filtering is used to remove accidental interference from single points and a small number of trajectories in the first clustering result, retaining only valid off-road clustering regions with statistical repeatability. Define G as the first clustering result generated in a single clustering, |G| as the total number of first off-road points (valid off-road points) within this first clustering result, and set a minimum valid sample threshold G. min =15.

[0152] The elimination decision formula is: |G| <G min .

[0153] If the number of points in the first cluster result satisfies the above elimination criterion formula, then the first cluster result is determined to be a set of random trajectories and is eliminated; only |G|≧G is retained. minThe first clustering result. This mechanism can effectively filter out interference samples that have no actual off-road significance, such as single vehicle mistrajectories, equipment positioning drift, and temporary road crossings, ensuring that the candidate regions have the repeatability of real behavior.

[0154] 3. Road Distance Check: Points with a center point less than 5 meters from the paved road are eliminated, thus ruling out non-off-road scenarios such as temporary roadside parking and driving along the road edge. This filter layer is based on accurate road network data and quickly filters points by calculating the Euclidean distance from the point to the nearest paved road. The 5-meter threshold effectively covers GPS positioning drift errors at the road edge while ensuring that all non-off-road activities within the roadway are eliminated.

[0155] 4. IQR Outlier Removal: Calculate the distance di from each point in the first cluster result to the cluster center. Calculate the interquartile range (IQR) using Q1 and Q3: IQR = Q3 - Q1. Remove outliers with distance di. <Q1-1.5 "IQR" or "di>Q3+1.5" The spatial outliers of "IQR" ensure that the spatial distribution of clusters is concentrated and continuous.

[0156] It should be noted that IQR (Interquartile Range) is a robust measure of dispersion. Compared to standard deviation, it is insensitive to extreme values ​​and is suitable for robustly identifying outliers in spatial data. Q1 is the first quartile (25th quartile), Q3 is the third quartile (75th quartile), and 1.5 times IQR is the classic Tukey outlier criterion.

[0157] 5. 5km POI check: Remove areas with more than 2 POIs within 5km of the cluster center, and exclude pseudo-off-road areas such as suburban villages, construction sites, and remote factory areas.

[0158] The 5-kilometer spatial range serves as a macroscopic assessment scale for the surrounding environment of the off-road area. The POI threshold of 2 is set according to the rule that if there are 3 or more artificial POIs within a 5-kilometer radius, it indicates that the area is located in a relatively active human zone and does not possess the "wilderness" attributes required for off-roading. This filtering layer complements the first layer's "POI + building + paved road strong filtering" in terms of spatial scale. The first layer verifies the micro-environment within the area, while this layer assesses the macro-environment surrounding the area.

[0159] 6. Off-road vehicle proportion filtering: This layer of filtering ensures that the main body of the candidate region trajectory is professional off-road vehicle, and removes false regions formed by interference from ordinary vehicles and non-off-road vehicles. For each first clustering result G, define N. off N represents the number of valid trajectories of off-road vehicles within the first clustering result. allThis represents the total number of valid trajectories for all vehicles within the first clustering result.

[0160] The formula for determining the purity of an off-road vehicle is: R off =N off / N all Set the purity threshold R. min =0.8. The exclusion criterion formula is: R off <R min .

[0161] When the proportion of off-road vehicle trajectories in the first cluster result is less than 80%, the region of the first cluster result is judged to have a large number of non-off-road vehicle behaviors mixed in with the trajectories, belonging to a low-confidence pseudo-off-road region, and the first cluster result is removed. By setting a purity threshold, it is possible to ensure that off-road vehicles occupy an absolute dominant position in the regional trajectories, effectively eliminating interference from areas where various types of vehicles mix and pass through, such as urban-rural fringe areas.

[0162] This filtering method retains only areas with high concentrations of off-road tracks (ensuring that the tracks in these areas are primarily from professional off-road vehicles), thus ensuring the accuracy of off-road area identification from a vehicle perspective.

[0163] 7. Trajectory Continuity Filtering: This layer of filtering distinguishes between continuous off-road driving and instantaneous trajectory deviation, eliminating fragmented points with no actual off-road behavior. Define T i Set a minimum effective off-road time threshold T for the continuous driving duration of a segment of a single effective off-road track. min =30s.

[0164] The elimination decision formula is: T i <T min .

[0165] The original trajectory segments corresponding to all first off-road checkpoints were traversed and verified, and the continuous driving time was less than T. min The trajectory is determined to be invalid, such as momentary deviation, positioning jitter, or temporary overtaking, and the corresponding first off-road checkpoint is removed. Only first off-road checkpoints with continuous trajectory and stable driving behavior are retained, ensuring that each first off-road checkpoint corresponds to real and continuous off-road driving behavior. By setting a minimum effective off-road time threshold, various accidental and momentary non-off-road activities can be effectively filtered out.

[0166] 8. Terrain Compliance Filtering: This layer filters standard off-road terrain based on terrain undulation features, eliminating flat areas such as plains and flatlands that have no off-road value. The maximum elevation within the candidate area is defined as H. max The minimum altitude is H min The difference in elevation between the topographic reliefs in the region is ΔH. ΔH = H max -H min .

[0167] Set the minimum off-road undulation threshold ΔH min =30m. The elimination criterion is: ΔH < ΔH min .

[0168] If the overall elevation difference of the region is less than ΔH min Areas deemed purely flat, artificial / natural flat land with no off-road terrain attributes are directly eliminated. Only complex terrain areas with significant undulations, such as mountains, Gobi Desert, desert, and riverbanks, are retained to match the terrain features of real off-road scenarios. By setting a minimum off-road undulation threshold, it is possible to effectively distinguish terrain areas with off-road value from flat areas lacking off-road characteristics.

[0169] 9. Secondary Clustering Noise Pre-filtering: This layer of filtering performs pre-noise reduction for the secondary accurate clustering, removing spatially isolated and non-clustering noise points in advance to improve the accuracy of subsequent clustering. Based on the spatial distribution characteristics within the first clustering result, the average neighborhood density of points within the first clustering result is defined as ρ, and the neighborhood density of a single point is defined as ρ0. i .

[0170] The formula for determining density deviation is: ρ i <0.5ρ.

[0171] Iterates through all points in the first clustering result. If the neighborhood density of a single point is significantly lower than the average neighborhood density, it is identified as a spatially isolated noise point, marked and removed in advance. This operation can remove remote discrete interference points in advance, avoid invalid noise points from participating in the secondary clustering, prevent problems such as false small clusters and clustering shifts, and maximize the accuracy and regional integrity of the secondary DBSCAN clustering.

[0172] The above is an explanation of the multi-level progressive filtering method. After completing the multi-level progressive filtering, all the second cross-country points that have been filtered and retained are stored in a memory cache with a simplified structure. Each record contains only the core information necessary for clustering calculation, such as the unique identifier ID and coordinates of the point, so that it can be efficiently read in subsequent accurate clustering steps, thereby minimizing the memory usage and I / O overhead of the secondary clustering input data.

[0173] In practical applications, the first clustering result obtained from loose clustering is subjected to the aforementioned nine-layer progressive filtering to obtain purified, discrete second cross-country points. Each filtering layer is executed serially, with each layer processing the data from the previous layer to ensure that the filtering effect accumulates step by step and continuously purifies the data.

[0174] Step S403: Perform precise clustering of the second off-road points to determine the off-road area.

[0175] It should be noted that precise clustering refers to a DBSCAN clustering configuration that uses a larger neighborhood radius (e.g., ε2 = 2000 meters) and a higher minimum point threshold (e.g., MinPts2 = 5).

[0176] In the specific implementation step S403, DBSCAN is used to perform precise clustering (secondary clustering) on ​​the second off-road points, thereby clustering these second off-road points into multiple independent, contiguous off-road areas with directly drawable boundaries.

[0177] In practical applications, the parameters for DBSCAN precise clustering can be set to "ε2=2000 meters, MinPts2=5", or the clustering parameters can be set according to the actual situation. The large neighborhood radius of ε2=2000 meters can aggregate spatially adjacent off-road points with a distance of less than 2 kilometers into the same off-road area, ensuring the spatial integrity of the off-road area; the minimum number of points threshold of MinPts2=5 requires that each off-road area contains at least 5 valid points to ensure that the area has a certain scale and confidence.

[0178] When performing DBSCAN precise clustering, it aggregates geographically continuous complete off-road areas by large neighborhoods, and uses a higher minimum point threshold to eliminate scattered interference points, thereby achieving precise aggregation of complete off-road areas.

[0179] After clustering, the minimum convex hull or circumscribed polygon of each cluster is extracted as the final spatial boundary of the off-road region. Simultaneously, attribute statistics are performed on each off-road region, including feature parameters such as region area, point density, number and proportion of off-road vehicles, and altitude range, providing complete regional attribute data for subsequent map labeling and information services.

[0180] The above embodiments of the present invention Figure 4 This section explains how to determine off-road areas. Firstly, the two-stage clustering strategy balances recall and precision, and only the second off-road point obtained after multi-layered progressive filtering can enter the precise cluster. The two-stage clustering method demonstrates good adaptability to off-road areas with varying spatial distribution densities.

[0181] Secondly, the serial architecture of loose clustering and multi-level progressive filtering controls computational overhead while ensuring recall. The first round of loose clustering aggregates spatially adjacent points, allowing subsequent filtering operations to be performed in batches on a group basis. Furthermore, the multi-level progressive filtering can be arranged in order of computational cost from low to high, and this sorting strategy optimizes the overall computational efficiency.

[0182] Thirdly, the off-road area boundaries output through precise clustering have a high degree of spatial consistency with the actual off-road activity range. Multi-layer progressive filtering has eliminated various pseudo-off-road scenarios such as interference from artificial facilities, accidental fragmented trajectories, mixed vehicle sources, and flat terrain. The points retained are all real off-road behavior records that have been verified and confirmed through multiple dimensions. Therefore, the off-road area boundaries output by this solution are closer to the spatial distribution of actual off-road behavior, avoiding the common problems of "large-scale mis-delineation" and "boundary omission" in geometric fitting methods. This provides more accurate geofencing data for subsequent downstream applications such as map annotation and navigation push.

[0183] After identifying the off-road area, the spatial boundary data of the off-road area can be further pushed to the vehicle map navigation system, off-road enthusiast community platform or outdoor sports APP, and visualized in the form of POI layer, heat map layer or vector boundary layer. At the same time, it supports searching and sorting by attributes such as area, activity level, terrain type, etc., providing off-road users with value-added services such as site discovery, route planning and safety warnings.

[0184] In summary, the above embodiments demonstrate that, firstly, this solution constructs a multi-dimensional, progressive hierarchical filtering system. Each layer specifically eliminates a typical misjudgment scenario, gradually converging candidate regions and addressing the high misjudgment rate of traditional single-dimensional identification, thus outputting highly reliable off-road areas. Secondly, this solution employs a two-stage adaptive DBSCAN clustering mechanism, balancing the prevention of missed identifications in sparse and remote off-road areas with the avoidance of excessive clustering in dense and popular areas, solving the industry problem of poor adaptability of traditional single-clustering parameters and significantly improving clustering robustness. Thirdly, this solution uses a vehicle whitelist classification system, accurately screening professional off-road vehicles based on vehicle hardware attributes, modification attributes, and historical driving trajectories, eliminating invalid trajectory data from ordinary family cars and commercial vehicles at the source, thereby reducing the identification misjudgment rate from the data layer.

[0185] Corresponding to the off-road area positioning method provided in the above embodiments of the present invention, see also... Figure 5 The present invention also provides a structural block diagram of a positioning system for off-road areas, the system comprising: a first filtering unit 501, a processing unit 502, a second filtering unit 503, and a clustering unit 504;

[0186] The first screening unit 501 is used to screen out the second driving trajectory representing off-road behavior from the first driving trajectories of multiple vehicles.

[0187] The processing unit 502 is used to identify the first off-road point from the second driving trajectory and determine the candidate area based on the first off-road point.

[0188] The second filtering unit 503 is used to filter out seed regions from the candidate regions.

[0189] Clustering unit 504 is used to perform multiple clusterings on the seed region to determine the off-road region.

[0190] In this embodiment of the invention, a second driving trajectory representing off-road behavior is selected, and then candidate regions are determined using the second driving trajectory. Seed regions are selected from the candidate regions, and multiple clustering operations are performed on the seed regions to determine the off-road regions. This method does not rely on preset geometric templates but identifies off-road regions based on off-road points, avoiding the erroneous introduction of a large number of non-off-road regions and improving the identification effect and positioning accuracy of off-road regions.

[0191] Preferred, combined Figure 5 The first filtering unit 501, as shown, includes an identification module, a verification module, and a first filtering module. The execution principle of each module is as follows:

[0192] The identification module is used to identify off-road vehicles from multiple vehicles by utilizing a pre-built whitelist of vehicles with off-road characteristics.

[0193] In some embodiments, the identification module is specifically used to: obtain multiple off-road determination conditions built into a pre-built whitelist of vehicles with off-road characteristics; and identify vehicles that meet any one of the off-road determination conditions from among multiple vehicles to identify off-road vehicles.

[0194] The verification module is used to verify the motion characteristics of the first driving trajectory of the off-road vehicle.

[0195] The first filtering module is used to filter out the second driving trajectory that represents the off-road behavior. The second driving trajectory is the first driving trajectory of the off-road vehicle that has passed the motion feature verification.

[0196] In this embodiment of the invention, a dual screening mechanism of vehicle whitelist and motion feature verification is used to compress the original second-by-second trajectory data of the Internet of Things, which greatly reduces the data scale of subsequent spatial calculations. At the same time, the serial architecture of dual verification ensures that misjudgments at each stage are intercepted in time, avoiding the cumulative error amplification effect of pseudo off-road trajectories on subsequent point identification and area aggregation. This ensures the accuracy and efficiency of the entire off-road area identification process from the source of data processing, and maximizes the elimination of invalid interference data without omitting real off-road trajectories.

[0197] Preferred, combined Figure 5 The processing unit 502, as shown, includes an extraction module, a conversion module, and a construction module. The execution principle of each module is as follows:

[0198] The extraction module is used to extract the center point of each trajectory segment in the second driving trajectory to obtain the first off-road point.

[0199] In some embodiments, the extraction module is specifically used to: slice and sample the second driving trajectory to obtain the trajectory center point of a single continuous trajectory segment; determine the trajectory center point of the single continuous trajectory segment as the first off-road point, the first off-road point including at least latitude and longitude coordinates, sampling time, trajectory duration, terrain elevation and distance from paved road.

[0200] The conversion module is used to perform coordinate projection conversion on the first off-road point.

[0201] The construction module is used to construct a preset area centered on the first cross-country point after coordinate projection transformation, and obtain candidate areas.

[0202] Preferred, combined Figure 5 The second filtering unit 503, as shown, includes a query module and a second filtering module. The execution principle of each module is as follows:

[0203] The query module is used to query the environmental attribute parameters of candidate regions.

[0204] Specifically, the query module is used to: call the map interface to query the preset feature data within the candidate area; and parse and classify the preset feature data to obtain the environmental attribute parameters of the candidate area.

[0205] The second filtering module is used to filter out seed regions from the candidate regions. The seed regions are the candidate regions whose environmental attribute parameters meet the preset filtering conditions.

[0206] Preferred, combined Figure 5 The clustering unit 504, as shown, includes a first clustering module, a filtering module, and a second clustering module. The execution principle of each module is as follows:

[0207] The first clustering module is used to perform loose clustering on the seed region to obtain the first clustering result.

[0208] The filtering module is used to perform multi-level progressive filtering on the first clustering results to obtain discrete second cross-country points.

[0209] The second clustering module is used to accurately cluster the second off-road points in order to determine the off-road area.

[0210] Preferably, the present invention also provides an electronic device, including: a processor and a memory, the processor and the memory being connected via a communication bus; wherein, the processor is used to call and execute a program stored in the memory; the memory is used to store the program, the program being used to implement the off-road area positioning method provided in the above method embodiments.

[0211] Preferably, embodiments of the present invention also provide a vehicle, the vehicle including the electronic equipment provided above.

[0212] In summary, the embodiments of the present invention provide a method, electronic device, and vehicle for locating off-road areas. The method filters out a second driving trajectory representing off-road behavior, and then uses this second driving trajectory to determine candidate areas. Seed areas are selected from the candidate areas, and these seed areas are clustered multiple times to determine the off-road areas. This method does not rely on preset geometric templates but identifies off-road areas based on off-road points, avoiding the erroneous introduction of a large number of non-off-road areas and improving the identification effect and positioning accuracy of off-road areas.

[0213] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0214] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0215] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for locating an off-road area, characterized in that, The method includes: Select the second driving trajectory representing off-road behavior from the first driving trajectories of multiple vehicles; Identify the first off-road point from the second driving trajectory, and determine the candidate area based on the first off-road point; Seed regions are selected from the candidate regions; The seed region is clustered multiple times to determine the off-road region.

2. The method according to claim 1, characterized in that, From the first driving trajectories of multiple vehicles, a second driving trajectory representing off-road behavior is selected, including: Using a pre-built whitelist of vehicles with off-road characteristics, off-road vehicles can be identified from multiple vehicles; The motion characteristics of the first driving trajectory of the off-road vehicle are verified. A second driving trajectory representing off-road behavior is selected, wherein the second driving trajectory is the first driving trajectory of the off-road vehicle verified by the motion characteristics.

3. The method according to claim 1, characterized in that, Identifying the first off-road point from the second driving trajectory, and determining candidate regions based on the first off-road point, including: Extract the center point of each trajectory segment in the second driving trajectory to obtain the first off-road point; Perform coordinate projection transformation on the first off-road point; A preset region is constructed centered on the first off-road point where the coordinate projection transformation has been completed, and a candidate region is obtained.

4. The method according to claim 1, characterized in that, Seed regions are selected from the candidate regions, including: Query the environmental attribute parameters of the candidate region; Seed regions are selected from the candidate regions, where the seed regions are the candidate regions whose environmental attribute parameters meet the preset filtering conditions.

5. The method according to claim 1, characterized in that, Multiple clustering operations are performed on the seed region to determine the off-road region, including: The seed region is subjected to loose clustering to obtain the first clustering result; The first clustering result is subjected to multi-layer progressive filtering to obtain discrete second off-road points; The second off-road location is precisely clustered to determine the off-road area.

6. The method according to claim 4, characterized in that, Query the environmental attribute parameters of the candidate region, including: Call the map interface to query the preset element data within the candidate area; The preset element data is parsed and classified statistically analyzed to obtain the environmental attribute parameters of the candidate region.

7. The method according to claim 2, characterized in that, Using a pre-built whitelist of vehicles with off-road characteristics, off-road vehicles are identified from multiple vehicles, including: Obtain multiple off-road determination conditions built into a pre-built whitelist of vehicles with off-road characteristics; Identify off-road vehicles from among multiple vehicles that meet any one of the aforementioned off-road determination criteria.

8. The method according to claim 3, characterized in that, Extract the center points of each trajectory segment in the second driving trajectory to obtain the first off-road point, including: The second driving trajectory is sliced ​​and sampled to obtain the trajectory center point of a single continuous trajectory segment; The center point of a single continuous trajectory segment is determined as the first off-road point. The first off-road point includes at least latitude and longitude coordinates, sampling time, trajectory duration, terrain elevation, and distance from paved road.

9. An electronic device, characterized in that, include: A processor and a memory are connected via a communication bus; wherein the processor is used to call and execute a program stored in the memory; The memory is used to store a program for implementing the off-road area positioning method as described in any one of claims 1 to 8.

10. A vehicle, characterized in that, The vehicle includes the electronic equipment as described in claim 9.

Images