Crowdsourcing mapping-based road change detection method and system, terminal and medium
By reconstructing and evaluating 3D models from vehicle perception data through crowdsourced mapping, the inaccuracy of high-precision map updates caused by data sparsity and lighting variations is solved, achieving stable and accurate updates of high-precision maps.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-03-24
AI Technical Summary
Existing technologies for crowdsourced mapping suffer from poor accuracy in updating high-precision maps due to issues such as data sparsity, occlusion, and lighting variations when assisted driving vehicles traverse complex road sections.
By continuously receiving vehicle perception data, a 3D model is reconstructed and evaluated. Only after the reconstruction quality passes the test can candidate change areas be identified by comparing with high-precision map tiles, and updates are triggered when preset conditions are met.
It improves the accuracy of high-precision map updates, reduces invalid updates caused by single false detections or temporary changes, and ensures the stability and freshness of map data.
Smart Images
Figure CN121725451A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data processing, and particularly relates to a road change detection method and device based on crowd-sourced mapping, a terminal and a medium. BACKGROUND
[0002] Traditional map updating processes rely on special collection vehicles for regular inspection, and map updating is completed by manual or semi-automatic processes. This approach has problems such as high deployment cost, long response period, limited coverage, and is difficult to meet the dual needs of real-time and wide coverage of map for mass-produced assisted driving systems.
[0003] In view of the above problems, in recent years, the idea of "crowd-sourced mapping" has been proposed, that is, relying on a large number of low-cost sensors (such as cameras, millimeter wave radars, and laser radars) deployed on assisted driving vehicles to collect environmental perception data, combining high-precision positioning information to reconstruct the local scene, and returning the perception results to the cloud to realize automatic detection and updating of map content. In the crowd-sourcing mode, the data sources are more diverse, the coverage is more extensive, and it has good scalability and economy, especially suitable for deployment of automatic driving systems for millions of vehicles.
[0004] However, in actual application, the time and space distribution of assisted driving vehicles through specific complex road sections has strong non-uniformity, and there may be problems such as data sparsity, occlusion, and light changes, which result in poor stability of single reconstruction results. The existing scheme often directly judges the candidate change area based on such reconstruction results with insufficient stability, which is easy to cause deviation in change recognition, and ultimately affects the accuracy of high-precision map updating.
[0005] Therefore, the prior art has defects and needs to be improved and developed. SUMMARY
[0006] The technical problem to be solved by the present application is to provide a road change detection method, system, terminal and medium based on crowd-sourced mapping, aiming at solving the problem of affecting the accuracy of high-precision map updating in the prior art.
[0007] The technical solution adopted by the present application to solve the technical problem is as follows: In a first aspect, the present application provides a road change detection method based on crowd-sourced mapping, comprising: continuously receiving perception data uploaded by a plurality of vehicles for a target road area; after the end of each collection period, reconstructing a current three-dimensional model of the target road area based on the perception data received in the current collection period; performing reconstruction quality evaluation on the current three-dimensional model, and if the reconstruction quality evaluation result is passed, comparing the current three-dimensional model with the corresponding high-definition map block to identify a candidate change region appearing in the current collection period; determining whether the candidate change region meets a preset update triggering condition, and triggering update of the high-definition map block if the candidate change region meets the preset update triggering condition.
[0008] In an implementation, the end of each collection period is triggered by a preset data amount condition, and the preset data amount condition is that the total amount of received perception data since the end of the last period reaches a capacity threshold or the coverage degree of newly added perception data to the target road region since the end of the last period reaches a coverage threshold.
[0009] In an implementation, the current three-dimensional model of the target road region is reconstructed based on the perception data received in the current collection period, including: desensitizing the perception data received in the current collection period to obtain desensitized data; performing spatio-temporal alignment and fusion on the desensitized data and eliminating dynamic objects to obtain fused perception data; constructing an initial three-dimensional model corresponding to the target road region based on the fused perception data; optimizing the initial three-dimensional model to generate the current three-dimensional model.
[0010] In an implementation, after performing reconstruction quality evaluation on the current three-dimensional model, the method further includes: if the reconstruction quality evaluation result is failed, suspending operation of the current three-dimensional model.
[0011] In an implementation, the comparison of the current three-dimensional model with the corresponding high-definition map block to identify the candidate change region appearing in the current collection period includes: registering and cutting the current three-dimensional model with the corresponding historical high-definition map; comparing the registered and cut current three-dimensional model with the high-definition map block, and regarding a region in which a geometric state has a significant change as a candidate change region.
[0012] In an implementation, the determination of whether the candidate change region meets the preset update triggering condition includes: obtaining identification records of the candidate change region in a preset number of continuous historical collection periods, the identification records including whether the candidate change region is identified in each period, a corresponding road change score and a change type; obtaining a road change score, a change type and geographical location information of the candidate change region in the current collection period; The comprehensive verification is performed based on all the identification records, the road change score and the change type corresponding to the current collection period, and the geographic location information, and when the comprehensive verification passes, it is determined that the preset update triggering condition is met.
[0013] In an embodiment, the comprehensive verification is performed based on all the identification records, the road change score and the change type corresponding to the current collection period, and the geographic location information, and includes: The first credibility verification is performed based on the road change score in all the identification records and the road change score corresponding to the current collection period. The second credibility verification is performed based on the change type in all the identification records and the change type corresponding to the current collection period. The stability verification is performed based on the number of times that the candidate change region is identified in all the identification records. The key region verification is performed based on the geographic location information of the candidate change region.
[0014] In a second aspect, an embodiment of the present application further provides a road change detection system based on crowd-sourced mapping, and the system includes: A data receiving module is configured to continuously receive perception data uploaded by a plurality of vehicles for a target road region. A reconstruction module is configured to, at the end of each collection period, reconstruct a current three-dimensional model of the target road region based on the perception data received in the current collection period. An identification module is configured to perform reconstruction quality evaluation on the current three-dimensional model, and if the reconstruction quality evaluation result is passed, the current three-dimensional model is compared with a corresponding high-definition map block to identify a candidate change region appearing in the current collection period. An update module is configured to judge whether the candidate change region meets a preset update triggering condition, and if so, trigger an update of the high-definition map block.
[0015] In a third aspect, an embodiment of the present application further provides a terminal, which includes a memory, a processor, and a road change detection program based on crowd-sourced mapping stored in the memory and executable on the processor, and when the road change detection program based on crowd-sourced mapping is executed by the processor, the steps of the road change detection method based on crowd-sourced mapping are implemented.
[0016] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, which stores a road change detection program based on crowd-sourced mapping, and the road change detection program based on crowd-sourced mapping can be executed to implement the steps of the road change detection method based on crowd-sourced mapping.
[0017] The beneficial effects of the present application are: the present application continuously receives perception data uploaded by multiple vehicles for a target road area; at the end of each collection period, a current three-dimensional model of the target road area is reconstructed based on the perception data received in the current collection period; the current three-dimensional model is evaluated for reconstruction quality, and if the reconstruction quality evaluation result is passed, the current three-dimensional model is compared with the corresponding high-precision map block to identify candidate change regions that appear in the current collection period; it is judged whether the candidate change region meets the preset update triggering condition, and if it meets, the update of the historical high-precision map is triggered. The present application evaluates the current three-dimensional model for reconstruction quality, and only when the evaluation is passed, the subsequent candidate change region identification and map update are performed, thereby effectively improving the accuracy of high-precision map update. BRIEF DESCRIPTION OF DRAWINGS
[0018] Figure 1 is a flowchart of a preferred embodiment of the road change detection method based on crowd-sourced mapping in the present application.
[0019] Figure 2 is a flowchart of the high-precision map block update in the present application.
[0020] Figure 3 is a structure diagram of a preferred embodiment of the road change detection system based on crowd-sourced mapping in the present application.
[0021] Figure 4 is a terminal principle block diagram of the present application. DETAILED DESCRIPTION
[0022] To make the purpose, technical scheme and advantages of the present application clearer and more explicit, the present application is further described in detail below with reference to the drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.
[0023] The traditional map update process relies on special collection vehicles for regular inspection, and the map update is completed by manual or semi-automatic process. This way has problems such as high deployment cost, long response period, limited coverage, etc., and it is difficult to meet the dual needs of real-time and wide coverage of the map for mass-produced assisted driving systems.
[0024] In view of the above problems, in recent years, the idea of "crowd-sourced mapping" has been proposed, that is, relying on a large number of low-cost sensors (such as cameras, millimeter wave radars, laser radars) deployed on assisted driving vehicles to collect environmental perception data, combining high-precision positioning information to reconstruct the local scene, and returning the perception results to the cloud to realize automatic detection and update of map content. In the crowd-sourcing mode, the data source is more abundant, the coverage area is more extensive, and it has good scalability and economy, especially suitable for deployment of automatic driving systems for millions of vehicles.
[0025] However, in practical applications, the spatiotemporal distribution of assisted driving vehicles traversing specific complex road sections exhibits strong non-uniformity, potentially leading to issues such as data sparsity, occlusion, and lighting variations, resulting in poor stability of single reconstruction results. Existing solutions often directly identify candidate change areas based on these unstable reconstruction results, which can easily cause biases in change identification and ultimately affect the accuracy of high-precision map updates.
[0026] To address the aforementioned deficiencies in existing technologies, this invention provides a road change detection method, system, terminal, and medium based on crowdsourced mapping. The method includes: continuously receiving perception data for a target road area uploaded by multiple vehicles; at the end of each acquisition cycle, reconstructing a current 3D model of the target road area based on the perception data received during the current acquisition cycle; evaluating the reconstruction quality of the current 3D model; if the reconstruction quality evaluation result is satisfactory, comparing the current 3D model with corresponding high-precision map tiles to identify candidate change areas appearing in the current acquisition cycle; determining whether the candidate change areas meet preset update trigger conditions; if so, triggering an update of the historical high-precision map. This invention effectively improves the accuracy of high-precision map updates by evaluating the reconstruction quality of the current 3D model and only proceeding with candidate change area identification and map updates after the evaluation is satisfactory.
[0027] Please see Figure 1 The road change detection method based on crowdsourced mapping according to an embodiment of the present invention includes the following steps: Step S100: Continuously receive perception data for the target road area uploaded by multiple vehicles.
[0028] Specifically, the vehicle of this invention is an assisted driving vehicle with driver assistance functions. Utilizing sensors on the assisted driving vehicle (such as cameras, millimeter-wave radar, lidar, etc.), environmental perception is performed on a target road area to obtain perception data. The perception data can be images, point clouds, etc. The target road area can be determined based on national road topology information and combined with business requirements (such as key monitoring areas, user activity hotspots, areas with frequent historical changes, etc.). This invention utilizes crowdsourced perception data from assisted driving vehicles to construct a map change detection system, replacing the traditional method of relying on dedicated data collection vehicles, significantly reducing the cost and deployment threshold of high-precision map updates, and increasing the frequency and breadth of update coverage.
[0029] Please see Figure 1 The road change detection method based on crowdsourced mapping described in this embodiment of the invention further includes the following steps: Step S200: After each acquisition cycle ends, the current three-dimensional model of the target road area is reconstructed based on the sensing data received during the current acquisition cycle.
[0030] Specifically, the end of each data collection cycle is triggered by a preset data volume condition. This preset condition is either that the total amount of received sensing data since the end of the previous cycle reaches a capacity threshold, or that the coverage of the target road area by newly added sensing data since the end of the previous cycle reaches a coverage threshold. After each data collection cycle, this invention specifically reconstructs the current 3D model of the target road area, enabling timely detection of changed areas and subsequent updates to the high-precision map.
[0031] In one implementation, based on the sensing data received within the current acquisition period, a current three-dimensional model of the target road area is reconstructed, including: The sensed data received during the current acquisition period is desensitized to obtain desensitized data; The desensitized data is spatiotemporally aligned and fused, and dynamic objects are removed to obtain fused perception data; An initial 3D model corresponding to the target road area is constructed based on the fused sensing data; The initial 3D model is optimized to generate the current 3D model.
[0032] Specifically, due to differences in sensor types, installation locations, and perception capabilities among different vehicles, spatiotemporal alignment and fusion are necessary. This approach significantly enhances the stability and integrity of reconstruction in complex road sections or data-sparse areas. By removing dynamic objects, subsequent identification is ensured to be based on the static environment, reducing misjudgments. The initial 3D model can be a 3D point cloud or a mesh model, without restriction. After generating the initial 3D model, bundle adjustment (BA) or pose graph optimization can be performed to eliminate accumulated errors, improve the global accuracy and consistency of the reconstructed model, and ultimately obtain the current 3D model.
[0033] Please see Figure 1 The road change detection method based on crowdsourced mapping described in this embodiment of the invention further includes the following steps: Step S300: Perform a reconstruction quality assessment on the current 3D model. If the reconstruction quality assessment result is satisfactory, compare the current 3D model with the corresponding high-precision map tiles to identify candidate change areas that appear in the current acquisition cycle.
[0034] Specifically, due to potential issues such as data sparsity, occlusion, and lighting variations in the sensor data, the amount of data collected in a certain area may be insufficient or of poor quality. This will affect the accuracy of subsequent processing in the generated 3D model. Therefore, a reconstruction quality assessment is necessary; only models that pass this assessment can proceed to the next step. The reconstruction quality assessment can include at least one of the following: point cloud density assessment, data viewpoint coverage assessment, and vehicle trajectory distribution assessment. For point cloud density assessment, it is necessary to determine whether the point cloud density exceeds a preset point cloud density threshold; only if it exceeds this threshold can the assessment be considered successful. For data viewpoint coverage assessment, it is necessary to determine whether the data viewpoint coverage exceeds a preset coverage threshold; only if it exceeds this threshold can the assessment be considered successful. For vehicle trajectory distribution assessment, it is necessary to determine whether the temporal distribution density of the vehicle trajectory exceeds a preset temporal density threshold; only if it exceeds this threshold can the assessment be considered successful. If the reconstruction quality assessment result is unsuccessful, operations on the current 3D model are paused, and the current 3D model that failed the quality assessment is deleted upon reaching a preset timeout. This method can improve the overall detection accuracy and efficiency, and avoid low-quality data affecting the judgment.
[0035] In one implementation, the current 3D model is compared with the corresponding high-precision map tiles to identify candidate change areas appearing in the current acquisition cycle, including: Register and crop the current 3D model with the corresponding historical high-precision map; The current 3D model after registration and cropping is compared with the high-precision map tiles, and regions where the geometric state has changed significantly are selected as candidate change regions.
[0036] Specifically, firstly, high-precision map tiles corresponding to the target road area are extracted from the high-precision map tile library, and the current 3D model is registered with these map tiles in a unified coordinate system. The ICP (Iterative ClosestPoint) algorithm is used during registration to achieve high-precision spatial alignment between the current 3D model and historical high-precision maps. Subsequently, concave hull models of both are constructed to determine their spatial extent, and intersection operations are performed on the concave hulls to obtain their common spatial boundary. Finally, the two high-precision map tiles are cropped based on this intersection concave hull to ensure that the data from both sides are within a completely consistent spatial region, thus providing a unified and accurate spatial basis for subsequent geometric difference detection and improving the accuracy of change detection. By comparing the registered and cropped current 3D model with the high-precision map tiles using methods such as distance field analysis and feature matching differences, areas with significant changes in geometric state (such as newly added or disappeared objects, changes in road surface shape, etc.) are identified as candidate change areas, achieving change identification at the road element level.
[0037] Please see Figure 1 The road change detection method based on crowdsourced mapping described in this embodiment of the invention further includes the following steps: Step S400: Determine whether the candidate change area meets the preset update trigger condition. If it does, trigger the update of the high-precision map block.
[0038] Specifically, the flowchart of high-precision map tile updating in this invention is as follows: Figure 2 As shown, after obtaining candidate change areas, the system does not immediately update the high-precision map. Instead, it conducts a comprehensive evaluation phase to determine whether the change is real, stable, and significant. This approach effectively avoids invalid updates caused by single false detections or temporary changes.
[0039] In one implementation, determining whether the candidate changed region meets a preset update trigger condition includes: The identification records of the candidate change area are obtained within a preset number of consecutive historical collection periods. The identification records include: whether the area was identified in each period, the corresponding road change score, and the change type. Obtain the road change score, change type, and geographical location information corresponding to the candidate change areas in the current collection period; Based on all the identified records, the road change score and change type corresponding to the current collection period, and the geographical location information, a comprehensive verification is performed. When the comprehensive verification passes, it is confirmed that the preset update triggering conditions are met.
[0040] Specifically, the system records the corresponding candidate change region each time it identifies a candidate change region. The domain is defined, and the corresponding road change score and change type are calculated.
[0041] The formula for calculating the road change score is: The road change score is a quantifiable measure of the degree of change, which can be used for comprehensive evaluation later.
[0042] The type of change can be determined based on perceptual detection of candidate change areas. Perceptual detection can utilize a trained deep learning model. Change types can include: new roadblocks / traffic cones, road marking repainting / wearing, construction site fencing, increase or decrease in the number of lanes, and relocation / content modification of traffic signs, etc.
[0043] The core of determining whether a candidate change region meets the preset update triggering conditions lies in performing a comprehensive multi-period, multi-dimensional verification of the candidate change region. The system traces the identification records of the region over multiple consecutive acquisition periods (including whether it was identified, road change scores, and change types), and performs a comprehensive verification based on all the identification records, the road change scores and change types corresponding to the current acquisition period, and geographical location information. Specific steps include: performing a first credibility verification based on the road change scores in all the identification records and the road change scores corresponding to the current acquisition period; performing a second credibility verification based on the change types in all the identification records and the change types corresponding to the current acquisition period; performing a stability verification based on the number of times the candidate change region was identified in all the identification records; and performing a key area verification based on the geographical location information of the candidate change region. The first credibility verification ensures that the observation results are reliable and have low noise, while the second credibility verification eliminates misjudgments caused by perceptual ambiguity. This allows the invention to not only focus on traditional geometric differences (such as the addition / disappearance of obstacles, road widening / narrowing, etc.) but also detect semantic-level changes (such as traffic sign relocation, temporary detours, lane marking changes, etc.), achieving comprehensive perception of structural and functional road changes. Stability verification ensures that changes are continuously identified over multiple periods, rather than appearing only briefly; critical area verification ensures priority is given to areas crucial to navigation safety, such as intersections and ramps. This invention does not rely on a single identification result but rather combines data from multiple collection periods for comprehensive judgment and introduces multiple verification methods. This approach effectively reduces invalid updates caused by false alarms or short-term fluctuations, ensuring that map data maintains both freshness and stability and accuracy.
[0044] In one implementation, when a high-precision map tile is updated, the cloud pushes the update to the vehicle terminal.
[0045] Specifically, this method achieves a closed-loop process of automatic detection and automatic update.
[0046] The method of this invention has strong scalability and platform independence, and can be adapted to different vehicle platforms, sensor combinations and perception algorithm outputs. It is suitable for various intelligent driving deployment scenarios, such as highways, urban roads and park roads.
[0047] In summary, this invention continuously receives perception data for a target road area uploaded by multiple vehicles. At the end of each acquisition cycle, based on the perception data received in the current acquisition cycle, it reconstructs the current 3D model of the target road area. The reconstruction quality of the current 3D model is evaluated; if the evaluation result is satisfactory, the current 3D model is compared with the corresponding high-precision map tiles to identify candidate change areas that appeared in the current acquisition cycle. It is then determined whether the candidate change areas meet preset update trigger conditions; if so, an update of the historical high-precision map is triggered. This invention effectively improves the accuracy of high-precision map updates by evaluating the reconstruction quality of the current 3D model and only proceeding with candidate change area identification and map updates after the evaluation is satisfactory.
[0048] In one embodiment, such as Figure 3 As shown, based on the above-described road change detection method based on crowdsourced mapping, the present invention also provides a road change detection system based on crowdsourced mapping, the system comprising: The data receiving module 100 is used to continuously receive perception data for a target road area uploaded by multiple vehicles; The reconstruction module 200 is used to reconstruct the current three-dimensional model of the target road area based on the sensing data received during the current acquisition cycle after the end of each acquisition cycle. The identification module 300 is used to evaluate the reconstruction quality of the current 3D model. If the reconstruction quality evaluation result is passed, the current 3D model is compared with the corresponding high-precision map block to identify the candidate change area that appears in the current acquisition cycle. The update module 400 is used to determine whether the candidate change area meets the preset update triggering conditions. If it does, the update of the high-precision map block is triggered.
[0049] In one embodiment, the system further includes: The desensitization unit is used to desensitize the sensed data received during the current acquisition cycle to obtain desensitized data. The fusion unit is used to perform spatiotemporal alignment and fusion of the desensitized data, and remove dynamic objects to obtain fused perception data; The construction unit is used to construct an initial three-dimensional model corresponding to the target road area based on the fused sensing data; An optimization unit is used to optimize the initial 3D model to generate the current 3D model.
[0050] In one embodiment, the system further includes: The pause unit is used to pause operations on the current 3D model if the reconstruction quality assessment result is unsatisfactory.
[0051] In one embodiment, the system further includes: The registration unit is used to register and crop the current 3D model with the corresponding historical high-precision map; The identification unit is used to compare the current 3D model after registration and cropping with the high-precision map tiles, and to identify regions where the geometric state has changed significantly as candidate change regions.
[0052] In one embodiment, the system further includes: The historical record acquisition unit is used to acquire the identification records of the candidate change area within a preset number of consecutive historical acquisition cycles. The identification records include: whether it was identified in each cycle, the corresponding road change score, and the change type. The current information acquisition unit is used to acquire the road change score, change type and geographical location information corresponding to the candidate change area in the current acquisition period; The comprehensive verification unit is used to perform comprehensive verification based on all the identification records, the road change score and change type corresponding to the current collection period, and the geographical location information. When the comprehensive verification passes, it confirms that the preset update triggering conditions are met.
[0053] In one embodiment, the system further includes: The first verification unit is used to perform a first confidence verification based on the road change scores in all the identification records and the road change scores corresponding to the current collection period. The second verification unit is used to perform a second confidence verification based on the change types in all the identified records and the change types corresponding to the current collection period. The third verification unit is used to perform stability verification based on the number of times the candidate change region is identified in all the identification records; The fourth verification unit is used to perform key area verification based on the geographical location information of the candidate change areas.
[0054] Based on the above embodiments, the present invention also provides a terminal, the structural schematic diagram of which is as follows: Figure 4As shown. The terminal includes a processor, memory, network interface, and display screen connected via a device bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating device and a road change detection program based on crowdsourced mapping. The internal memory provides an environment for the operation of the operating device and the road change detection program based on crowdsourced mapping in the non-volatile storage medium. The network interface is used for communication with external terminals via a network connection. When executed by the processor, the road change detection program based on crowdsourced mapping implements the steps of any of the aforementioned road change detection methods based on crowdsourced mapping. The display screen can be a liquid crystal display (LCD) or an e-ink display.
[0055] Those skilled in the art will understand that Figure 4 The structural schematic diagram shown is only a schematic diagram of a part of the structure related to the present invention solution, and does not constitute a limitation on the terminal on which the present invention solution is applied. The specific terminal may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.
[0056] In one embodiment, a terminal is provided, the terminal including a memory, a processor, and a road change detection program based on crowdsourced mapping stored in the memory and executable on the processor. When the road change detection program based on crowdsourced mapping is executed by the processor, it implements the steps of any road change detection method based on crowdsourced mapping provided in the embodiments of the present invention.
[0057] This invention also provides a computer-readable storage medium storing a road change detection program based on crowdsourced mapping. When executed by a processor, the road change detection program based on crowdsourced mapping implements the steps of any of the road change detection methods based on crowdsourced mapping provided in this invention.
[0058] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0059] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0060] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0061] Those skilled in the art will 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, or a combination of computer software and electronic hardware. 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.
[0062] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units described above is only a logical functional division, and in actual implementation, it can be divided in other ways. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.
[0063] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not mean that the essence of the corresponding technical solutions deviates from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A road change detection method based on crowdsourced mapping, characterized in that, The method includes: It continuously receives perception data for the target road area uploaded by multiple vehicles; After each acquisition cycle ends, the current three-dimensional model of the target road area is reconstructed based on the sensing data received during the current acquisition cycle. The reconstruction quality of the current 3D model is evaluated. If the reconstruction quality evaluation result is passed, the current 3D model is compared with the corresponding high-precision map tiles to identify candidate change areas that appear in the current acquisition cycle. Determine whether the candidate changed area meets the preset update trigger condition. If it does, then trigger the update of the high-precision map block.
2. The road change detection method based on crowdsourced mapping according to claim 1, characterized in that, The end of each collection cycle is triggered by a preset data volume condition, which is either the total amount of received sensing data since the end of the previous cycle reaches a capacity threshold or the coverage of the target road area by newly added sensing data since the end of the previous cycle reaches a coverage threshold.
3. The road change detection method based on crowdsourced mapping according to claim 1, characterized in that, Based on the sensing data received during the current acquisition period, reconstruct the current 3D model of the target road area: The sensed data received during the current acquisition period is desensitized to obtain desensitized data; The desensitized data is spatiotemporally aligned and fused, and dynamic objects are removed to obtain fused perception data; An initial 3D model corresponding to the target road area is constructed based on the fused sensing data; The initial 3D model is optimized to generate the current 3D model.
4. The road change detection method based on crowdsourced mapping according to claim 1, characterized in that, After assessing the reconstruction quality of the current 3D model, the process also includes: If the reconstruction quality assessment result is unsatisfactory, operations on the current 3D model will be suspended.
5. The road change detection method based on crowdsourced mapping according to claim 1, characterized in that, The current 3D model is compared with the corresponding high-precision map tiles to identify candidate change areas that appear in the current acquisition cycle, including: Register and crop the current 3D model with the corresponding historical high-precision map; The current 3D model after registration and cropping is compared with the high-precision map tiles, and regions where the geometric state has changed significantly are selected as candidate change regions.
6. The road change detection method based on crowdsourced mapping according to claim 1, characterized in that, Determining whether the candidate changed region meets the preset update triggering condition includes: The identification records of the candidate change area are obtained within a preset number of consecutive historical collection periods. The identification records include: whether the area was identified in each period, the corresponding road change score, and the change type. Obtain the road change score, change type, and geographical location information corresponding to the candidate change areas in the current collection period; Based on all the identified records, the road change score and change type corresponding to the current collection period, and the geographical location information, a comprehensive verification is performed. When the comprehensive verification passes, it is confirmed that the preset update triggering conditions are met.
7. The road change detection method based on crowdsourced mapping according to claim 6, characterized in that, A comprehensive verification is performed based on all the aforementioned identification records, the road change score and change type corresponding to the current collection period, and geographical location information, including: Based on the road change scores in all the aforementioned identification records and the road change scores corresponding to the current collection period, a first confidence verification is performed; A second confidence verification is performed based on the change types in all the identified records and the change types corresponding to the current collection period; Stability is verified based on the number of times the candidate change region is identified in all the identification records. Based on the geographical location information of the candidate change areas, key areas are verified.
8. A road change detection system based on crowdsourced mapping, characterized in that, include: The data receiving module is used to continuously receive perception data for the target road area uploaded by multiple vehicles; The reconstruction module is used to reconstruct the current three-dimensional model of the target road area based on the sensing data received during the current acquisition cycle after the end of each acquisition cycle. The identification module is used to evaluate the reconstruction quality of the current 3D model. If the reconstruction quality evaluation result is passed, the current 3D model is compared with the corresponding high-precision map tiles to identify the candidate change areas that appear in the current acquisition cycle. The update module is used to determine whether the candidate change area meets the preset update triggering conditions. If it does, the update of the high-precision map block is triggered.
9. A terminal, characterized in that, The terminal includes: a memory, a processor, and a road change detection program based on crowdsourced mapping stored in the memory and executable on the processor. When the road change detection program based on crowdsourced mapping is executed by the processor, it implements the steps of the road change detection method based on crowdsourced mapping as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a road change detection program based on crowdsourced mapping, which, when executed by a processor, implements the steps of the road change detection method based on crowdsourced mapping as described in any one of claims 1-7.