Global relocation method and device and vehicle

Abstract

The invention discloses a global relocation method and device and a vehicle. The method comprises the steps of obtaining a target scene where the vehicle is currently located; determining scene parameters corresponding to the target scene based on the target scene; determining a target search range of the vehicle according to the scene parameters, a set position search range and a set angle search range; based on the target search range, position and angle sampling is carried out under the prior poses of the vehicle, and a plurality of scattered prior poses are obtained; and determining a relocation result of the vehicle based on the plurality of scattered prior poses. According to the method, the possibility of matching failure can be effectively reduced, and the relocation success rate of the vehicle in any scene is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to a global repositioning method, device, and vehicle. Background Art

[0002] As the popularity of smart cars continues to accelerate, users' requirements for diversified functions of smart cars have increased dramatically, especially the smart driving function in the parking lot, which is basically one of the must-have functions of smart cars nowadays. Among them, functions such as P2P assisted parking have become the biggest selling point of new energy vehicles.

[0003] The success rate of launching functions in any scenario will directly reflect the level of intelligent driving of automobile manufacturers. The main reason affecting the launch of functions in any scenario is the success rate of vehicle repositioning in any scenario. Due to inaccurate pre-positioning of vehicles in scenarios such as underground parking lots, occluded scenes, and weak real-time dynamic differential positioning technology (RTK), initial positioning fails, resulting in the inability to launch intelligent driving functions. Therefore, to improve the launch rate of intelligent driving functions, it is equivalent to improving the success rate of vehicle repositioning in any scenario, but the existing technology still has the problem of low repositioning success rate. Summary of the Invention

[0004] The embodiments of the present application provide a global relocation method, device, and vehicle. The method utilizes dynamic expansion to relocate the search range, which can effectively reduce the possibility of matching failure, improve robustness, and effectively improve the success rate and accuracy of vehicle relocation in any scenario.

[0005] In the first aspect, the present application provides the following technical solutions through an embodiment of the present application:

[0006] A global relocalization method includes: obtaining a target scene in which a vehicle is currently located; determining scene parameters corresponding to the target scene based on the target scene; determining a target search range for the vehicle based on the scene parameters, a set position search range, and a set angle search range; sampling positions and angles based on the target search range under a priori posture of the vehicle to obtain multiple scattered priori postures; and determining a relocalization result for the vehicle based on the multiple scattered priori postures.

[0007] Optionally, the target scene includes an open scene, and obtaining the target scene in which the vehicle is currently located includes: if the absolute position information of the vehicle and the memory position information of the vehicle when it was last powered off are obtained, then determining whether the absolute position information is a fixed solution, wherein the absolute position information is obtained based on the inertial measurement unit information, speed information and satellite positioning information of the vehicle; if it is a fixed solution, then determining whether the distance between the absolute position information and the memory position information is less than a distance threshold; if so, determining that the target scene in which the vehicle is currently located is an open scene.

[0008] Optionally, the method also includes: if the distance between the absolute posture information and the memory posture is greater than or equal to the distance threshold, performing the following scene recognition steps: obtaining image data collected by the vehicle; bringing the image data into a preset scene recognition model to obtain the scene type and confidence results of the vehicle, wherein the scene recognition model is trained based on different image data, different scene types and confidence levels; based on the scene type and confidence results, determining the target scene in which the vehicle is currently located.

[0009] Optionally, before obtaining the memory posture information of the vehicle when it was last powered off, it also includes: reading the memory posture information pre-stored in the vehicle; if the reading is successful, obtaining the memory posture information of the vehicle when it was last powered off; if the reading fails, executing the scene recognition step.

[0010] Optionally, before sampling the position and angle under the prior posture of the vehicle based on the target search range, it also includes: obtaining the prior posture of the vehicle; obtaining the prior posture of the vehicle includes: if the memory posture information is read successfully, then determining the memory posture information as the prior posture of the vehicle; if the memory posture information fails to be read, then determining the absolute posture information as the prior posture of the vehicle.

[0011] Optionally, the target scene in which the vehicle is currently located is determined based on the result of the scene type and the confidence level, including: obtaining a status value and a search star number in the satellite positioning information; determining a result coefficient under the scene type based on the status value and the search star number; updating the confidence level under the scene type based on the result coefficient; judging whether the updated confidence level is greater than a set confidence level, and if so, determining that the scene type is an available scene; and determining the target scene in which the vehicle is currently located based on the available scene.

[0012] Optionally, determining the target scene in which the vehicle is currently located based on the available scenes includes: determining whether the scene types in the scene recognition results of N consecutive frames are all the available scenes; if so, determining that the scene type is the target scene in which the vehicle is currently located; if not, determining that the scene type of the last successful recognition is the target scene.

[0013] Optionally, determining the relocation result of the vehicle based on the multiple scattered prior postures includes: performing semantic matching on the multiple scattered prior postures in sequence to obtain a matching evaluation score and posture information corresponding to each prior posture; if the matching evaluation score corresponding to a certain prior posture exceeds a set score, determining that the semantic matching of the corresponding prior posture is successful; and determining the relocation result of the vehicle based on the number of prior postures that have successfully matched semantically, multiple matching evaluation scores that have successfully matched semantically, and multiple posture information.

[0014] In a second aspect, the present application provides the following technical solutions through an embodiment of the present application:

[0015] A global relocation device, comprising:

[0016] The acquisition module is used to obtain the target scene where the vehicle is currently located;

[0017] A parameter determination module, configured to determine, based on the target scene, scene parameters corresponding to the target scene;

[0018] A range determination module, configured to determine a target search range for the vehicle based on the scene parameters, a set position search range, and a set angle search range;

[0019] A posture determination module is used to sample the position and angle of the vehicle under the prior posture based on the target search range to obtain multiple scattered prior postures;

[0020] A positioning module is used to determine a repositioning result of the vehicle based on the multiple scattered prior poses.

[0021] In a third aspect, the present application provides the following technical solutions through an embodiment of the present application:

[0022] A vehicle comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods described in the first aspect when executing the program.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] The global relocation method provided in the embodiment of the present application first determines the target scene in which the vehicle is currently located, and according to the target scene, clarifies the scene parameters for revising the set position search range and the set angle search range. The scene parameters are combined with the position search range and the angle search range to determine the target search range of the vehicle in the target scene. Then, the position and angle are sampled under the prior posture of the vehicle to obtain multiple scattered prior postures. Finally, based on the multiple scattered prior postures, the relocation result of the vehicle is determined. The present application classifies the vehicle operation scenes. The search range and search angle of different target scenes are different. The search mode is dynamically adjusted to dynamically expand the relocation search area, which is conducive to improving the search efficiency and the reliability of positioning, and enhancing the relevance of the search results. It can effectively reduce the occurrence of matching failures, improve the success rate of relocation, and improve the robustness of the system, thereby effectively improving the success rate and accuracy of vehicle relocation in any scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application, a brief introduction will be given below to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 This is a flowchart of the global relocation method in an embodiment of the present application;

[0027] Figure 2 This is a schematic structural diagram of a global relocation device in an embodiment of the present application;

[0028] Figure 3 This is a schematic diagram of the structure of the vehicle in the embodiment of the present application. DETAILED DESCRIPTION

[0029] The embodiments of the present application provide a global relocation method, device, and vehicle. The method utilizes dynamic expansion to relocate the search range, which can effectively reduce the possibility of matching failure, improve robustness, and effectively improve the success rate and accuracy of vehicle relocation in any scenario.

[0030] The technical solution of the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0031] A global relocalization method comprises: obtaining a target scene in which a vehicle is currently located; determining scene parameters corresponding to the target scene based on the target scene; determining a target search range of the vehicle according to the scene parameters, a set position search range, and a set angle search range; sampling positions and angles under a priori posture of the vehicle based on the target search range to obtain multiple scattered priori postures; and determining a relocalization result of the vehicle based on the multiple scattered priori postures.

[0032] In order to better understand the above technical solution, the following will be combined with the accompanying drawings and specific implementation methods to describe the above technical solution in detail. In the first aspect, the embodiment of the present application provides a global relocation method, specifically, Figure 1 As shown, the method includes the following steps S101 to S105:

[0033] Step S101, obtaining the target scene where the vehicle is currently located.

[0034] It should be noted that the prerequisite for executing the global repositioning method in this application may be: executing the method when the vehicle is powered on or when the vehicle positioning confidence is low, and the condition for ending the global repositioning method may be: ending the positioning when the vehicle positioning confidence is normal.

[0035] In a specific embodiment, the target scene may include an open scene, and obtaining the target scene in which the vehicle is currently located may include: if the absolute position information of the vehicle and the memory position information of the vehicle when it was last powered off are obtained, then determining whether the absolute position information is a fixed solution, wherein the absolute position information is obtained based on the vehicle's inertial measurement unit information, speed information and satellite positioning information; if it is a fixed solution, then determining whether the distance between the absolute position information and the memory position information is less than a distance threshold; if so, determining that the target scene in which the vehicle is currently located is an open scene.

[0036] In one application scenario, before obtaining the target scene in which the vehicle is currently located, it can also include: obtaining the vehicle's Inertial Measurement Unit (IMU) information, speed information, satellite positioning (Global Navigation Satellite System, GNSS) information and image data at the current moment, and converting the IMU information, speed information, GNSS information and image data into the data format required by the software, wherein the GNSS information includes the GNSS posture Pgnss.

[0037] Furthermore, in order to improve the accuracy of data acquisition, after obtaining the IMU information, speed information and GNSS information, it can also include: preprocessing the data; the preprocessing includes: determining whether there is a break in the IMU information, speed information, GNSS information and image data, and determining whether there is an abnormality in the IMU information, speed information, GNSS information and image data.

[0038] Specifically, first determine whether there is a flow interruption in the above four information. The method of determining the flow interruption is, for example: if it is detected that the next information is not obtained within a specified period of time, it is determined that there is a flow interruption. If there is a flow interruption in one of the information, the fault is reported to the system to control the vehicle to slow down and stop on the side of the road, and continue to execute the steps of obtaining IMU information, speed information, GNSS information and image data.

[0039] If there is no interruption in the information flow, the step of determining whether there is an abnormality in the information is performed: determine whether there is an abnormality in the above four information respectively. The judgment method is as follows: if a certain information is detected as irrelevant data, the information is determined to be abnormal and the fault is reported to the system to control the vehicle to slow down and stop, and continue to obtain IMU information, speed information, GNSS information and image data. If there is no abnormality in the information, the IMU information, speed information and GNSS information are fused to obtain the absolute position information P global .

[0040] As an embodiment, an IMU installed on the vehicle may be used to obtain IMU information, a GNSS posture monitor installed on the vehicle may be used to obtain GNSS information, and a camera installed on the vehicle may be used to obtain image data of the vehicle's external environment.

[0041] In one application scenario, before obtaining the vehicle's last power-off stored posture information, the process may also include: reading the stored posture information pre-stored in the vehicle; if the reading is successful, obtaining the vehicle's last power-off stored posture information; if the reading fails, executing the scene recognition step. For example, the stored posture information may be information stored in a specified memory after the vehicle is engaged in P gear, for subsequent retrieval. The reason for the failed reading may be file loss or corruption, for example.

[0042] In one embodiment, if the memory posture information is read successfully, the memory posture flag F of the vehicle is adjusted. isStoredPose Set to true; if reading fails, adjust the memory posture flag F isStoredPose Set to false,

[0043] If the absolute position information of the vehicle and the memory position information of the vehicle when it was last powered off are obtained, that is, if the memory position flag F is determined isStoredPoseIf the result is true, execute the next step, including: judging whether the absolute posture information is a fixed solution, if it is a fixed solution, then judging the memory posture P stored With the absolute pose P global Is the distance between them within the distance threshold D? stored_th According to the judgment result, the flag F is determined isInArea .

[0044] In a specific embodiment, the flag F is determined based on the distance between the two. isInArea Methods may include:

[0045]

[0046] If the flag F is determined isInArea If the result is true, adjust the scene recognition start flag F open_scene Is false; if the flag F is determined isInArea If the result is false, adjust the scene recognition start flag F open_scene Optionally, the distance threshold can be between 1.5 and 2.5 meters, for example, 2 meters.

[0047] Next, determine the scene recognition start flag F open_scene If the result is false, the scene is determined to be an open scene S open .

[0048] This application uses a method of determining the deviation between the absolute posture information and the memory posture information to eliminate situations where the vehicle has moved significantly, i.e., the deviation is large and the vehicle has a large displacement, such as the vehicle being towed, which makes the memory posture information read at the last power-off have a significant disadvantage. When it is determined that the distance between the memory posture and the absolute posture is less than a distance threshold, the vehicle's current target scene is determined to be an open scene.

[0049] The target scene may also include an underground garage scene, an occluded scene, and a semi-occluded scene. The method further includes: if the distance between the absolute posture information and the memory posture is greater than or equal to a distance threshold, executing the following scene recognition steps: acquiring image data collected by the vehicle; bringing the image data into a preset scene recognition model to obtain the scene type and confidence results of the vehicle, wherein the scene recognition model is trained based on different image data, different scene types, and confidence levels; and determining the target scene in which the vehicle is currently located based on the scene type and confidence results.

[0050] In a specific embodiment, if it is determined that the scene recognition start flag F open_sceneIf the result is true, or the memory posture information reading fails, the scene recognition step is executed: the image data P at the current moment (such as the i-th moment) is obtained. img_i Passed to the scene recognition model, output scene type and confidence result C i (S type_i ,S std_i ), scene type S type ∈{S open ,S half_hide ,S hide ,S underground_parking}, confidence S std_i ∈[0,1], where the scene recognition model is pre-trained based on different image data, different scene types and confidence levels. Different image data corresponds to different scene types and confidence levels.

[0051] Of course, as other optional embodiments, the target scenario may also include other scenarios that exist in actual use, such as: weak signal scenario, rooftop parking scenario, etc.

[0052] In one embodiment, in order to filter out erroneous results calculated by the model, verify the model calculation results, and improve the accuracy of the model calculation, the target scene in which the vehicle is currently located is determined based on the results of the scene type and the confidence level, which may include: obtaining the status value and the number of searched stars in the satellite positioning information; determining the result coefficient under the scene type based on the status value and the number of searched stars; updating the confidence level under the scene type based on the result coefficient; judging whether the updated confidence level is greater than the set confidence level, and if so, determining that the scene type is an available scene; and determining the target scene in which the vehicle is currently located based on the available scene.

[0053] In a specific embodiment, the model calculation result C at the i-th moment i Verification is performed, including: obtaining the state G in the GNSS data at the i-th moment of step statis_i and search star number G sateNum_i , calculation result coefficient

[0054] Where W i In the normalization process, the maximum value is 50, and the judgment coefficient W is required. i , if W i If W is greater than 1, i Set to 1; if W i Less than or equal to 1, keep W i constant.

[0055] Update the confidence S of the model output std_i =S std_i ×W i, judge whether the confidence is greater than the set confidence S std_th Optionally, the confidence level can be set between 0.4 and 0.6, for example, 0.5. If the confidence level is greater than the set confidence level, the scene type is considered to be an available scene. If the confidence level is less than the set confidence level, the scene type is considered to be an unavailable scene. Based on the determined available scene, the target scene in which the vehicle is currently located is determined.

[0056] Furthermore, in order to improve the accuracy of recognition, the target scene in which the vehicle is currently located is determined based on the available scenes, which may include: determining whether a certain scene type is an available scene in the scene recognition results of N consecutive frames; if so, determining that the scene type is the scene type that was successfully recognized, as the target scene in which the vehicle is currently located; if not, determining that the scene type that was successfully recognized last time is the target scene.

[0057] In a specific embodiment, determining whether a certain scene type in the scene recognition results of N consecutive frames is an available scene may include: obtaining image data collected by the vehicle at the next moment (i+1), substituting the image data into a preset scene recognition model, and obtaining the scene type and confidence level of the vehicle; based on the scene type and confidence level, determining whether the scene type is an available scene; then obtaining image data collected at the next moment (i+2), substituting the image data into a preset scene recognition model, and obtaining the scene type and confidence level of the vehicle; based on the scene type and confidence level, determining whether the scene type is an available scene; repeating the aforementioned steps until determining whether the scene type corresponding to the image data at the i+N-1th moment is an available scene, thereby obtaining scene recognition results for N consecutive frames. Optionally, N frames may be 6 to 10 frames, for example, 8 frames.

[0058] It should be noted that if the scene recognition result at the i-th moment is currently being determined, the scene recognition results for N consecutive frames can be directly obtained based on the results at the i-N+1 to i-th moments.

[0059] For example, if the scene types of the present application include open scenes, underground garage scenes, occluded scenes, and semi-occluded scenes, by acquiring image data at multiple consecutive moments, for the image data at a single moment, the image data is brought into the scene recognition model to obtain the scene type and confidence results at that moment, and after obtaining the scene type (for example, open scene) and confidence (for example, 1) results, the results are verified, and the confidence output of the model is updated according to the verification results; it is determined whether the updated confidence is greater than the set threshold. If it is greater than the set threshold, the result is determined to be available once (the open scene is an available scene), and the underground garage scene, occluded scene, and semi-occluded scene are determined to be unavailable scenes; if it is less than or equal to the set threshold, the result is determined to be unavailable (the open scene is an unavailable scene), and the underground garage scene, occluded scene, and semi-occluded scene are also determined to be unavailable scenes. The same steps are performed on the data at each moment to obtain the judgment results of the four scenes.

[0060] Normalize the judgment results of the above four scenarios, such as converting the available scenario to 1 and the unavailable scenario to 0, and then set the N (N=8) bit variable Q to open , Q half_hide , Q half , Q underground_parking In the above example, determine whether the values ​​of the four variables are equal to 2. N If the value of -1 is present, the current scene recognition is considered reliable and the recognized scene result is output; if not, the scene type of the last successful recognition is used. This method determines whether the results of 8 consecutive frames are all the same scene, and then determines whether the scene recognition result is reliable. If the result of a frame is a different scene, it is considered unreliable and the last recognition scene needs to be used to improve the recognition result of the optimized model.

[0061] This application uses N consecutive frames of scene recognition results to self-check the current model recognition results, reducing the probability of model result recognition errors and greatly improving the reliability of scene type classification.

[0062] In one embodiment, if the results of more than M identifications are unreliable, the system will not be initialized and uploaded to the system for fault reporting, so that the vehicle system control function is activated to improve safety. For example, M is 20 to 30 times.

[0063] Step S102: Based on the target scene, determine scene parameters corresponding to the target scene.

[0064] In a specific embodiment, it is assumed that the target scenes include open scenes, underground garage scenes, blocked scenes and semi-blocked scenes. The scene parameters corresponding to the open scenes, semi-blocked scenes, blocked scenes and underground garage scenes are a, b, c and d, respectively, where a is less than or equal to d, b is greater than a and d, and c is greater than b.

[0065] In one embodiment, the target scene is an open scene S open , the corresponding scene parameters are determined to be 1 to 1.5; the target scene is a semi-occluded scene S half_hide , the corresponding scene parameters are determined to be 2 to 3; the target scene is the occlusion scene S hide , the corresponding scene parameters are determined to be 3 to 5; the target scene is an underground garage scene, and the corresponding scene parameters are determined to be 1 to 1.5.

[0066] For example, according to the target scene, the scene type S type_i Mapped into the corresponding scene parameter D i :

[0067] S open →1

[0068] S half_hide →2

[0069] S hide →4

[0070] S Underground_parking →1

[0071] Step S103, determining a target search range for the vehicle according to the scene parameters, the set position search range, and the set angle search range;

[0072] Step S104, based on the target search range, sampling the position and angle under the prior posture of the vehicle to obtain multiple scattered prior postures;

[0073] Step S105 : determining a relocation result of the vehicle based on the multiple scattered prior poses.

[0074] In a specific embodiment, the scene parameter is multiplied by the set position search range to obtain an updated position search range, and the scene parameter is multiplied by the set angle search range to obtain an updated angle search range. By modifying the search range and angle based on the target scene, the recognition area and recognition method are dynamically adjusted to improve the reliability of search results, accelerate search efficiency, and reduce operating power consumption.

[0075] Specifically, according to the scene parameter D i Update the location search range d at time i const and the angle search range θconst , get the target search range, including:

[0076] d i =D i ×d const

[0077] θ i =D i ×θ const

[0078] For example, the position search range may be set to be between 4 and 6 meters, and the angle search range may be set to be between 5 and 10 degrees.

[0079] In a specific embodiment, based on the target search range, before sampling the position and angle under the vehicle's prior posture, the method may further include: obtaining the vehicle's prior posture; the method of obtaining the vehicle's prior posture includes: if the memory posture information is read successfully, that is, F isStoredPose If true, the memory posture information is determined as the vehicle's prior posture; if the memory posture information fails to be read, F isStoredPose If it is false, the absolute pose information is determined as the vehicle's prior pose.

[0080] Then, the updated position at time i is used to search the range d i and angle search range θ i , in the prior pose P pre Nearby (d i ,θ i ) uniformly sample positions and angles to obtain k scattered prior poses where j∈[1,k].

[0081] In one embodiment, determining a vehicle relocation result based on multiple scattered prior postures may include: semantically matching the multiple scattered prior postures in sequence to obtain a matching evaluation score and posture information corresponding to each prior posture; if the matching evaluation score corresponding to a certain prior posture exceeds a set score, determining that the semantic matching of the corresponding prior posture is successful; and determining a vehicle relocation result based on the number of prior postures with successful semantic matching, the total matching evaluation score of successful semantic matching, the single matching evaluation score of successful semantic matching, and the posture information.

[0082] Specifically, the k prior poses are semantically matched in sequence to obtain the matching evaluation score j and posture Judge k scores separately j Whether it exceeds the set score th(Depending on the matching algorithm), if it exceeds the set score, it is considered that the semantic matching under this prior posture is successful; if it is less than or equal to the set score, it is considered that the semantic matching under this prior posture fails.

[0083] In a specific embodiment, based on the number of prior poses with successful semantic matching, multiple matching evaluation scores for successful semantic matching, and multiple pose information, determining a relocalization result of the vehicle may include:

[0084] The multiple matching evaluation scores of successful semantic matches are summed up to obtain the total matching evaluation score of successful semantic matches; the vehicle relocation result is determined based on the number of prior poses of successful semantic matches, the total matching evaluation score of successful semantic matches, the multiple matching evaluation scores of successful semantic matches, and the multiple pose information.

[0085] The k matching results are judged. Assuming that h of them meet the requirements, the final relocation result is the weighted sum of the h matching results. The specific calculation formula is as follows:

[0086]

[0087]

[0088] Get the vehicle relocation result P final , where score all is the total score of successful semantic matching, and h is the number of prior poses with successful semantic matching. This weighted strategy, which combines the matching results with the scores, not only improves the accuracy of relocalization but also enhances its stability, ensuring the success rate of vehicle relocalization in any scenario.

[0089] The global relocalization method proposed in this application includes data preprocessing, judgment of memory posture information, scene recognition and relocalization steps. It uses a scene recognition model to classify vehicle operation scenarios, improves the compatibility of relocalization with rich scenarios, and uses dynamic expansion to relocalize the search range, which can effectively reduce the occurrence of matching failures and improve system robustness. It uses N consecutive frames of scene recognition results to self-check the current model recognition results, reduces the probability of model result recognition errors, and greatly improves the reliability of scene type classification. Finally, a weighted strategy that uses the matching results and scores for weighting not only improves the accuracy of relocalization, but also improves the stability of relocalization, ensures the success rate of vehicle relocalization in any scenario, not only improves the success rate of function activation in any scenario, but also improves the comfort and safety of user experience.

[0090] The beneficial effects brought by this method include at least: using the scene recognition model to identify the operating scene, thereby improving the dynamic scalability of the repositioning search area; the post-processing logic of the model recognition results can effectively improve the repositioning success rate and efficiency; using the memory pose as the prior pose, which can effectively reduce the positioning accuracy of the system under weak GNSS; using the continuous multi-frame judgment scene recognition to self-check the current scene recognition result strategy, which can effectively improve the correctness of the scene type; using the dynamic search area strategy, which can effectively improve the repositioning success rate.

[0091] To sum up, a global relocation method provided by an embodiment of the present application classifies vehicle operation scenarios, and the search ranges of different target scenarios are different. The dynamic adjustment of the search range is adopted to dynamically expand the relocation search area, which is conducive to enhancing the relevance of search results and improving search efficiency. It can effectively reduce the occurrence of matching failures, increase the success rate of relocation, and enhance the robustness of the system, thereby effectively improving the success rate and accuracy of vehicle relocation in any scenario.

[0092] In the second aspect, based on the same inventive concept, the embodiment of the present application provides a global relocation device, such as Figure 2 As shown, including:

[0093] An acquisition module 401 is used to acquire the target scene in which the vehicle is currently located;

[0094] A parameter determination module 402 is configured to determine, based on the target scene, scene parameters corresponding to the target scene;

[0095] A range determination module 403 is configured to determine a target search range for the vehicle based on the scene parameters, a set position search range, and a set angle search range;

[0096] A posture determination module 404 is configured to sample positions and angles of the vehicle at the prior posture based on the target search range to obtain a plurality of scattered prior postures;

[0097] The positioning module 405 is configured to determine a repositioning result of the vehicle based on the plurality of scattered priori postures.

[0098] As an optional embodiment, the target scene includes an open scene, and the acquisition module 401 is specifically used to: if the absolute position information of the vehicle and the memory position information of the vehicle when it was last powered off are obtained, then determine whether the absolute position information is a fixed solution, wherein the absolute position information is obtained based on the vehicle's inertial measurement unit information, speed information and satellite positioning information; if it is a fixed solution, then determine whether the distance between the absolute position information and the memory position information is less than a distance threshold; if so, determine that the target scene currently located by the vehicle is an open scene.

[0099] As an optional embodiment, the acquisition module 401 is specifically used to: if the distance between the absolute posture information and the memory posture is greater than or equal to the distance threshold, then perform the following scene recognition steps: obtain image data collected by the vehicle; bring the image data into a preset scene recognition model to obtain the scene type and confidence results of the vehicle, wherein the scene recognition model is trained based on different image data, different scene types and confidence levels; based on the scene type and confidence results, determine the target scene in which the vehicle is currently located.

[0100] As an optional embodiment, the acquisition module 401 is specifically used to: read the memory posture information pre-stored in the vehicle; if the reading is successful, obtain the memory posture information of the vehicle when it was last powered off; if the reading fails, execute the scene recognition step.

[0101] As an optional embodiment, the device also includes a priori posture determination module, which is used to: obtain the priori posture of the vehicle; obtain the priori posture of the vehicle, including: if the memory posture information is read successfully, then determine the memory posture information as the priori posture of the vehicle; if the memory posture information is read unsuccessfully, then determine the absolute posture information as the priori posture of the vehicle.

[0102] As an optional embodiment, the acquisition module 401 is specifically used to: obtain the status value and the number of searched stars in the satellite positioning information; determine the result coefficient under the scene type based on the status value and the number of searched stars; update the confidence under the scene type based on the result coefficient; determine whether the updated confidence is greater than the set confidence, and if so, determine that the scene type is an available scene; and determine the target scene in which the vehicle is currently located based on the available scene.

[0103] As an optional embodiment, the acquisition module 401 is specifically used to: determine whether the scene types in the scene recognition results of N consecutive frames are all available scenes; if so, determine that the scene type is the target scene where the vehicle is currently located; if not, determine that the scene type successfully recognized last time is the target scene.

[0104] As an optional embodiment, the positioning module 405 is specifically used to: perform semantic matching on multiple scattered prior postures in sequence to obtain the matching evaluation score and posture information corresponding to each prior posture; if the matching evaluation score corresponding to a certain prior posture exceeds the set score, it is determined that the semantic matching of the corresponding prior posture is successful; based on the number of prior postures with successful semantic matching, multiple matching evaluation scores of successful semantic matching, and multiple posture information, the vehicle repositioning result is determined.

[0105] The above modules can be implemented by software codes, in which case the above modules can be stored in the memory of the control device. The above modules can also be implemented by hardware such as integrated circuit chips.

[0106] The global relocation device provided in the embodiment of the present application has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference can be made to the corresponding content in the aforementioned method embodiment.

[0107] In a third aspect, based on the same inventive concept, the embodiment of the present application provides a vehicle 500, such as Figure 3 As shown, it includes: a memory 501, a processor 502 and a computer program 503 stored in the memory and capable of running on the processor. When the processor 502 executes the program, the steps of the global relocation method described in the first aspect are implemented.

[0108] Since the vehicle described in the embodiment of the present application is the vehicle used for implementing the global relocation method in the embodiment of the present application, based on the global relocation method described in the embodiment of the present application, those skilled in the art will be able to understand the specific implementation of the vehicle in the embodiment of the present application and its various variations, so how to implement the method in the embodiment of the present application for this vehicle will not be described in detail here. As long as those skilled in the art implement the vehicle used in the global relocation method in the embodiment of the present application, it falls within the scope of protection to be provided by this application.

[0109] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0110] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A module that specifies functions in one or more boxes.

[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction module, which is implemented in the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0113] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0114] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

Claims

1. A global relocation method, characterized in that: include: Get the target scene the vehicle is currently in; Based on the target scene, determining scene parameters corresponding to the target scene; Determining a target search range for the vehicle based on the scene parameters, a set position search range, and a set angle search range; Based on the target search range, sampling the position and angle under the prior posture of the vehicle to obtain multiple scattered prior postures; A relocalization result of the vehicle is determined based on the multiple scattered prior poses.

2. The method according to claim 1, wherein The target scene includes an open scene, and obtaining the target scene where the vehicle is currently located includes: If absolute position information of the vehicle and memorized position information of the vehicle when it was last powered off are obtained, determining whether the absolute position information is a fixed solution, wherein the absolute position information is obtained based on inertial measurement unit information, speed information, and satellite positioning information of the vehicle; If it is a fixed solution, it is determined whether the distance between the absolute posture information and the memory posture information is less than a distance threshold. If it is less than, it is determined that the target scene currently located by the vehicle is an open scene.

3. The method according to claim 2, wherein Also includes: If the distance between the absolute posture information and the memory posture information is greater than or equal to the distance threshold, the following scene recognition steps are performed: Obtain image data collected by the vehicle; Submitting the image data to a preset scene recognition model to obtain a result of the scene type and confidence level of the vehicle, wherein the scene recognition model is trained based on different image data, different scene types, and confidence levels; Based on the results of the scene type and the confidence level, the target scene in which the vehicle is currently located is determined.

4. The method according to claim 3, wherein If the vehicle's last power-off memory posture information is obtained before the vehicle is powered off, the following steps are also included: Read the memory posture information pre-stored in the vehicle; If the reading is successful, the vehicle's memory posture information at the time of last power-off is obtained; If the reading fails, the scene recognition step is performed.

5. The method according to claim 4, wherein Before sampling the position and angle of the vehicle based on the target search range and the prior posture of the vehicle, the method further includes: Obtaining a priori pose of the vehicle; The obtaining of the vehicle's prior posture includes: if the memory posture information is read successfully, determining the memory posture information as the vehicle's prior posture; if the memory posture information is read unsuccessfully, determining the absolute posture information as the vehicle's prior posture.

6. The method according to claim 3, wherein Determining the target scene in which the vehicle is currently located based on the result of the scene type and the confidence level includes: Obtaining the status value and the number of searched satellites in the satellite positioning information; Determining a result coefficient under the scene type according to the state value and the number of searched stars; Based on the result coefficient, updating the confidence level under the scenario type; Determining whether the updated confidence level is greater than a set confidence level, and if so, determining that the scene type is an available scene; According to the available scenes, a target scene in which the vehicle is currently located is determined.

7. The method according to claim 6, wherein Determining the target scene in which the vehicle is currently located based on the available scenes includes: Determining whether, in scene recognition results of consecutive N frames, the scene types are all the available scenes; If so, determining that the scene type is the target scene in which the vehicle is currently located; If not, it is determined that the scene type successfully identified last time is the target scene.

8. The method according to claim 1, wherein Determining a relocalization result of the vehicle based on the multiple scattered prior poses includes: Sequentially perform semantic matching on the multiple scattered prior postures to obtain a matching evaluation score and posture information corresponding to each prior posture; If the matching evaluation score corresponding to a certain priori posture exceeds the set score, the semantic matching of the corresponding priori posture is determined to be successful; A relocalization result of the vehicle is determined based on the number of prior poses that have successfully matched the semantics, multiple matching evaluation scores of the successful semantic matches, and multiple pose information.

9. A global relocation device, characterized in that: include: The acquisition module is used to obtain the target scene where the vehicle is currently located; A parameter determination module, configured to determine, based on the target scene, scene parameters corresponding to the target scene; A range determination module, configured to determine a target search range for the vehicle based on the scene parameters, a set position search range, and a set angle search range; A posture determination module is used to sample the position and angle of the vehicle under the prior posture based on the target search range to obtain multiple scattered prior postures; A positioning module is used to determine a repositioning result of the vehicle based on the multiple scattered prior poses.

10. A vehicle, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the method according to any one of claims 1 to 8 are implemented when the processor executes the program.

Images