Map construction method and device based on public source data, equipment, medium and product

Abstract

The invention provides a map construction method and device based on crowd source data, equipment, a medium and a product, and relates to the field of high-precision maps, and the method comprises the steps: obtaining a vehicle end mapping result which is transmitted by a crowd source vehicle end and is collected for multiple times, the vehicle end mapping result is a single-trip mapping result obtained after the public source vehicle end collects map sensing data and carries out vectorization processing on the map sensing data; in response to aggregated data obtained by performing multi-pass data aggregation on the single-pass mapping result, obtaining element data corresponding to each map element according to observation results obtained by performing multi-pass observation on the same map element in the aggregated data, the element data being used for generating fused mapping data during association fusion; and according to a matching relationship between existing map data and the fused mapping data, carrying out geometric correction on the fused mapping data so as to construct a public source map. By means of the method, the map building efficiency and precision can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of high-precision map technology, and in particular to a method, device, equipment, medium and product for constructing a map based on crowd-source data. Background Art

[0002] At present, the construction and updating method of high-precision maps is mainly through the acquisition of data by professional collection vehicles equipped with single or multiple high-precision lidars, and the collected data is transmitted to the map production line to complete the mapping work, which is costly and inefficient.

[0003] In related technologies, professional data collection vehicles are gradually replaced by crowd-source data collection to reduce the data collection cost of professional data collection vehicles. However, due to the continuous photos and images required for map construction, the volume of data transmitted back from the crowd-source vehicle is extremely large, the transmission cost is too high and there are high requirements for storage space. In addition, in scenarios with poor positioning signals such as multi-layer roads, parallel roads or in tunnels, it is easy to have problems with incorrect mapping results due to insufficient trajectory accuracy of crowd-source data. Summary of the invention

[0004] The present application provides a method, device, equipment, medium and product for constructing a map based on crowd-source data to solve at least one of the above-mentioned technical problems.

[0005] According to one aspect of the present application, a map construction method based on crowd-source data is provided, which is applied to the cloud, and includes: according to a pre-triggered mapping task, obtaining a vehicle-side mapping result sent by a crowd-source vehicle-side after multiple collections, the vehicle-side mapping result is a single-pass mapping result obtained after the crowd-source vehicle-side collects map perception data and vectorizes the map perception data; in response to aggregated data obtained by performing multiple data aggregation on the single-pass mapping result, obtaining element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, the element data being used to generate fused map data during association fusion; and according to a matching relationship between pre-stored existing map data and the fused map data, geometrically correcting the fused map data to construct a crowd-source map.

[0006] In one embodiment, the method also includes: acquiring trajectory data transmitted by the crowd-source vehicle-side in real time, and matching the trajectory data with preset trajectory data used for mapping to obtain a trajectory matching result; acquiring the vehicle-side mapping result sent by the crowd-source vehicle-side after multiple rounds of collection, including: when the trajectory matching result is consistent, triggering a mapping task to the crowd-source vehicle-side, the mapping task is used to instruct the crowd-source vehicle-side to send the vehicle-side mapping result, so as to obtain the vehicle-side mapping result sent by the crowd-source vehicle-side after multiple rounds of collection.

[0007] In one embodiment, the obtaining of element data corresponding to each map element based on the observation results of multiple observations of the same map element in the aggregated data includes: performing element fusion processing on the same map element based on the observation results of multiple observations of the same map element in the aggregated data to obtain fusion results of each map element, the fusion results including fusion position information and / or fusion geometric information; determining the position information and / or geometric information corresponding to each map element based on the fusion position information and / or the fusion geometric information, and obtaining the element data of each map element based on the position information and / or geometric information.

[0008] In one embodiment, the method further includes: performing registration and classification on each map element in the aggregated data based on road topology information of existing map data to obtain a registration and classification result, wherein the registration and classification result is used to indicate the correspondence between the same map elements at the same position in the aggregated data; performing element fusion processing on the same map element based on the observation results of multiple observations of the same map element in the aggregated data, including: performing element fusion processing on the same map element based on the registration and classification result and the observation result.

[0009] In one embodiment, after constructing the crowd-source map, the method further includes: for each map element in the crowd-source map, obtaining a consistency recognition result between the map element and the corresponding map element in the aggregated data, and at least one of the positioning accuracies corresponding to the map element; calculating the confidence of each map element in the crowd-source map and at least one of the positioning accuracies corresponding to the map element based on the consistency recognition result of each map element; performing data differentiation on the map elements having a confidence higher than a confidence threshold and the corresponding map elements in the latest version of the high-precision map data, and updating the corresponding map elements in the high-precision map data based on the differentiation result.

[0010] According to a second aspect of the present application, a map construction method based on crowd-source data is provided, which is applied to a crowd-source vehicle side, including: collecting map perception data during driving, and vectorizing the map perception data to obtain a single-pass mapping result; based on a multi-pass collection process in the above steps, obtaining a vehicle-side mapping result after multiple passes of collection; sending the vehicle-side mapping result after multiple passes of collection to the cloud, so that the cloud side constructs a crowd-source map based on the vehicle-side mapping result.

[0011] In one embodiment, the vectorizing the map perception data to obtain a single-pass mapping result includes: mapping the map perception data to an absolute coordinate system according to the posture information and calibration parameters of the multi-source vehicle end to obtain coordinate-converted perception data; performing motion compensation on the perception data according to a preset motion distortion equation, and tracking map elements of previous and next frames on the motion-compensated perception data to obtain tracked multi-frame perception data; and vectorizing the multi-frame perception results to obtain a single-pass mapping result.

[0012] In one embodiment, the processing of tracking map elements of the previous and next frames on the motion-compensated perception data includes: extracting map element features in the motion-compensated perception data according to a preset feature extraction algorithm; associating the map element features between the previous and next frames to obtain an association result of the map element features; and performing map element tracking processing on the previous and next frames on the motion-compensated perception data according to the association result.

[0013] In one embodiment, the vectorizing the multi-frame perception results to obtain a single-pass mapping result includes: converting the format of the map elements in the multi-frame perception results into a vector format to obtain vectorized map elements; constructing a topological structure based on the adjacency relationship between the map elements in the multi-frame perception results; and constructing a single-pass map based on the vectorized map elements and the topological structure to obtain the single-pass mapping result.

[0014] According to a third aspect of the present application, a map construction device based on crowd-source data is provided, which is applied to the cloud, including: a first acquisition module, which is configured to obtain the vehicle-side mapping results sent by the crowd-source vehicle-side after multiple collections according to a pre-triggered mapping task, the vehicle-side mapping results are single-pass mapping results obtained after the crowd-source vehicle-side collects map perception data and vectorizes the map perception data; a second acquisition module, which is configured to respond to aggregated data obtained by performing multiple data aggregation on the single-pass mapping results, and obtain element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, the element data being used to generate fused map data during associated fusion; a first mapping module, which is configured to perform geometric correction on the fused map data according to a matching relationship between pre-stored existing map data and the fused map data, so as to construct a crowd-source map.

[0015] In one embodiment, the device also includes: a trajectory acquisition module, which is configured to acquire the trajectory data transmitted by the multi-source vehicle end in real time, and match the trajectory data with the preset trajectory data used for mapping to obtain a trajectory matching result; the first acquisition module is specifically configured to: when the trajectory matching result is consistent, trigger a mapping task to the multi-source vehicle end, and the mapping task is used to instruct the multi-source vehicle end to send the vehicle-end mapping result, so as to obtain the vehicle-end mapping result sent by the multi-source vehicle end after multiple rounds of collection.

[0016] In one embodiment, the second acquisition module includes: a fusion unit, which is configured to perform element fusion processing on the same map element based on the observation results of multiple observations of the same map element in the aggregated data, to obtain a fusion result of each map element, wherein the fusion result includes fusion position information and / or fusion geometric information; a determination unit, which is configured to determine the position information and / or geometric information corresponding to each map element based on the fusion position information and / or the fusion geometric information, and obtain the element data of each map element based on the position information and / or geometric information.

[0017] In one embodiment, the device also includes: a registration and classification module, which is configured to perform registration and classification on each map element in the aggregated data based on the road topology information of the existing map data to obtain a registration and classification result, and the registration and classification result is used to indicate the correspondence between the same map elements at the same position in the aggregated data; the fusion unit is specifically configured to perform element fusion processing on the same map element based on the registration and classification result and the observation result.

[0018] In one embodiment, the device also includes: an identification module, which is configured to obtain, for each map element in the crowd-source map, a consistency identification result between the map element and the corresponding map element in the aggregated data, and at least one of the positioning accuracies corresponding to the map element; a confidence calculation module, which is configured to calculate the confidence of each map element in the crowd-source map according to the consistency identification result of each map element and at least one of the positioning accuracies corresponding to the map element; a differential update module, which is configured to perform data differentiation between the map elements whose confidence is higher than the confidence threshold and the corresponding map elements in the latest version of the high-precision map data, and update the corresponding map elements in the high-precision map data according to the differential result.

[0019] According to a fourth aspect of the present application, a map construction device based on crowd-source data is provided, which is applied to a crowd-source vehicle side, including: a second mapping module, which is configured to collect map perception data during driving, and vectorize the map perception data to obtain a single-pass mapping result; a third acquisition module, which is configured to obtain a vehicle-side mapping result after multiple passes of collection based on a multiple-pass collection process of the second mapping module; a transmission module, which is configured to send the vehicle-side mapping result after multiple passes of collection to the cloud, so that the cloud can construct a crowd-source map based on the vehicle-side mapping result.

[0020] In one embodiment, the second mapping module includes: a coordinate conversion unit, which is configured to map the map perception data to an absolute coordinate system according to the posture information and calibration parameters of the multi-source vehicle end to obtain the perception data after coordinate conversion; a compensation tracking unit, which is configured to perform motion compensation on the perception data according to a preset motion distortion equation, and perform map element tracking processing on the previous and next frames on the motion-compensated perception data to obtain tracked multi-frame perception data; a vectorization processing unit, which is configured to perform vectorization processing on the multi-frame perception results to obtain a single-pass mapping result.

[0021] In one embodiment, the compensation tracking unit is specifically configured to: extract map element features from the motion-compensated perception data according to a preset feature extraction algorithm; associate the map element features between previous and next frames to obtain association results of the map element features; and perform map element tracking processing on the motion-compensated perception data with respect to previous and next frames according to the association results.

[0022] In one embodiment, the vectorization processing unit is specifically configured to: convert the format of the map elements in the multi-frame perception results into a vector format to obtain vectorized map elements; construct a topological structure based on the adjacency relationship between the map elements in the multi-frame perception results; and construct a single-pass map based on the vectorized map elements and the topological structure to obtain the single-pass mapping result.

[0023] According to a fifth aspect of the present application, an electronic device is provided, comprising: a memory and a processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, so that the electronic device executes the map construction method based on crowd-source data.

[0024] According to a sixth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the map construction method based on crowd-source data.

[0025] According to a seventh aspect of the present application, a computer program product is provided, the computer program product comprising a computer program, which implements the map construction method based on crowd-source data when executed by a processor.

[0026] According to an eighth aspect of the present application, a map construction system based on crowd-source data is provided, comprising a cloud, and a crowd-source vehicle terminal electrically connected to the cloud; the cloud is used to execute the map construction method based on crowd-source data provided in any one of the first aspects above; the crowd-source vehicle terminal is used to execute the map construction method based on crowd-source data provided in any one of the second aspects above.

[0027] The map construction method, device, equipment, medium and product based on crowdsource data provided by this application, by performing single-trip mapping during the driving process of the vehicle, adopting the vehicle-side mapping method, only needs to transmit back the vector data after the vehicle-side mapping, which greatly reduces the amount of data transmission, relieves the cloud computing power and improves the mapping efficiency. The cloud observes the aggregated data corresponding to the vehicle-side mapping results collected by the vehicle for multiple trips multiple times, so as to obtain more accurate map element data based on the observation results and fuse them into a map, and correct the fused map in combination with existing map data (such as navigation maps, high-precision map data), so as to supplement the information of some scenes with poor positioning signals, thereby optimizing the mapping effect and improving the mapping accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0029] Figure 1 A possible scenario diagram provided for an embodiment of the present application;

[0030] Figure 2 A flowchart of a method for constructing a map based on crowdsource data provided in an embodiment of the present application;

[0031] Figure 3 A flowchart of another method for constructing a map based on crowdsource data provided in an embodiment of the present application;

[0032] Figure 4 yes Figure 3 Schematic diagram of the process of step S202;

[0033] Figure 5 It is a flowchart of another method for constructing a map based on crowdsource data provided in an embodiment of the present application;

[0034] Figure 6 It is a flowchart of another method for constructing a map based on crowd-source data provided in an embodiment of the present application;

[0035] Figure 7a yes Figure 6 One of the flowchart diagrams of step S601;

[0036] Figure 7b yes Figure 6 2. Flow chart of step S601 in FIG.

[0037] Figure 8 It is one of the interaction diagrams between the cloud and the multi-source vehicle end in the embodiment of the present application;

[0038] Fig. 9 This is the second interactive diagram between the cloud and the multi-source vehicle end in the embodiment of the present application;

[0039] Fig.10 is a structural schematic diagram of a map construction device based on crowd-source data provided in an embodiment of the present application;

[0040] Fig.11 is a structural schematic diagram of another map construction device based on crowd-source data provided in an embodiment of the present application;

[0041] Fig.12 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0042] Fig.13 A schematic diagram of the structure of another electronic device provided in an embodiment of the present application;

[0043] Fig.14 A schematic diagram of the structure of a map construction system based on crowd-source data provided in an embodiment of the present application;

[0044] Fig.15 It is a block diagram of a terminal device provided by an exemplary embodiment of the present application.

[0045] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0046] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0047] There are currently two ways to produce and update high-precision maps. One is to obtain data through a professional high-precision data collection vehicle equipped with a single or multiple high-precision laser radars, and then complete the data mapping and updating work through the map production line with the participation of a large number of people; the other is a high-precision map update method based on crowd-source forms, which collects photos or images from the vehicle and transmits them back to the cloud, where mapping is done. The above two methods have the following disadvantages:

[0048] For the first method, the high-precision data collection vehicle is equipped with a large number of high-precision professional equipment, which leads to high costs, often ranging from 60 to 1 million. The high price leads to a limited number of vehicles, which cannot support large-scale high-freshness maintenance of high-precision maps. For the second method, compared with the first method mentioned above, crowd-source mapping can use cheap equipment, usually no more than 1,000 yuan for a single hardware device, which greatly reduces the cost of mapping. However, the current crowd-source mapping method usually sends the photos and images collected by the vehicle back to the cloud for mapping. Due to the continuous photos and images required for mapping, the volume of data sent back is very large, resulting in high traffic costs, and the requirements for cloud storage space are also extremely high. In addition, in the current crowd-source mapping method, the data available on the vehicle side is usually the vehicle trajectory, photos or videos collected by the forward or surround view camera. In scenes with poor positioning signals such as multi-layer roads, parallel roads or tunnels, the mapping results may be incorrect due to insufficient trajectory accuracy, resulting in low mapping accuracy.

[0049] In view of this, the embodiment of the present application provides a map construction solution based on crowdsource data, which performs single-trip mapping during the driving process of the vehicle, adopts the vehicle-side mapping method, and only needs to transmit back the vector data after the vehicle-side mapping, which greatly reduces the amount of data transmission, relieves the cloud computing power and improves the mapping efficiency. The cloud aggregates the vehicle-side mapping results collected by the vehicle through multiple trips and observes the aggregated data multiple times to obtain more accurate map element data based on the observation results for fusion mapping, and can correct the fused mapping in combination with existing map data (such as navigation maps, high-precision map data), so as to supplement the information of some scenes with poor positioning signals, greatly optimize the mapping effect, and improve the mapping accuracy.

[0050] The embodiments of the present application are explained below in conjunction with application scenarios. The map construction based on crowd-source data provided by the embodiments of the present application can be applied to the application scenario of intelligent driving, and more specifically, can be applied to the application scenario of autonomous driving based on vehicle cloud computing. For example, the execution subject of the method provided by the embodiments of the present application can be the cloud or the crowd-source vehicle side. In some embodiments, the corresponding execution subject of the cloud can also be a map production platform based on a physical server. The technical solution of this application does not specifically limit the specific execution subject. The following introduces the cloud and the crowd-source vehicle side as the execution subjects of the method provided by the embodiments of the present application.

[0051] Figure 1 A schematic diagram of a scenario of a method for constructing a map based on crowdsource data provided in an embodiment of the present application, such as Figure 1 As shown, the crowd-source vehicle end 110 and the cloud end 120 are electrically connected. It can be understood that the cloud end can be a cloud server, and the crowd-source vehicle end can be a vehicle that provides crowd-source services, which can be an operating or personal vehicle. The crowd-source visual device is installed in front or in the back of the vehicle, which can realize the single-trip mapping of the vehicle end. In this embodiment, the transmission of crowd-source data (i.e., the vehicle-end mapping results) of two vehicles is used as an example for explanation. In actual applications, it can include more or fewer interactions between vehicles and the cloud, that is, the vehicle-end mapping results collected by more vehicles (or one vehicle) for multiple trips on the same road section. Among them, the crowd-source vehicle end 110 uses photos or videos collected by the forward-looking or surround-view camera during driving, and realizes the perception of road elements of single-frame images on the vehicle end through perception algorithms (such as AI image recognition technology), which can identify key elements including ground markings, arrows, text, traffic lights, traffic signs, etc. Then, by fusing and vectorizing the multi-frame perception results, the vehicle-side single-trip mapping result is obtained. After multiple-trip collection, the crowd-source vehicle-side 110 transmits the vehicle-side mapping result collected multiple times to the cloud 120. The cloud 120 aggregates the multi-trip data of the vehicle-side and observes the aggregated data multiple times. It uses the observation results to obtain map feature objects with higher credibility and higher accuracy, and fuses these map elements into a map. It also establishes a matching relationship between the existing navigation / high-precision map data and the fused mapping data, uses the high-precision map geometry data to align and correct the mapping geometry, and supplements the broken geometry, so as to increase the accuracy and correctness of the mapping data.

[0052] The following is a detailed description of the technical solution of the present application and how the technical solution of the present application solves the above technical problems in conjunction with the accompanying drawings and specific embodiments. It should be noted that these specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0053] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0054] Figure 2 A flowchart of a method for constructing a map based on crowdsource data provided in an embodiment of the present application, which is applied to the cloud, may include the following steps S201-S203:

[0055] Step S201, according to the pre-triggered mapping task, obtain the vehicle-side mapping result sent by the multi-source vehicle-side after multiple collections, wherein the vehicle-side mapping result is a single-pass mapping result obtained after the multi-source vehicle-side collects map perception data and vectorizes the map perception data.

[0056] In related technologies, the crowd-source mapping method usually uses crowd-source vehicles to collect road image data or impact data, and directly transmits these images or impact data to the cloud for mapping. The amount of data transmitted is very large, which places extremely high demands on the computing power and storage capacity of the cloud.

[0057] In this embodiment, when a mapping task is triggered, a single-trip mapping is performed on the vehicle side. The multi-source visual equipment installed in front or behind the vehicle is used to collect image data during the driving process of the vehicle and collect perception data through the perception algorithm to realize the perception of road elements in a single frame image. Key elements including ground markings, arrows, text, traffic lights, traffic signs, etc. can be identified. Then, by fusing the multi-frame perception results to obtain fused perception data, the fused perception data is vectorized to obtain the single-trip mapping result of the vehicle side. The multi-source vehicle side can obtain more detailed information and improve accuracy by performing multiple collections on the same road section. Among them, the triggering process of the mapping task can be realized by comparing the trajectory data uploaded by the multi-source vehicle side in the cloud. The specific process has been introduced in the following embodiments and will not be repeated here. In some embodiments, the mapping task can also be triggered by the user interacting with the cloud or other methods. Based on this method, the efficiency of vehicle-side mapping can be effectively improved to provide users with the map products they need.

[0058] By building maps on the vehicle side, there is no need to transmit large amounts of data back to the cloud (the amount of data can be reduced to 1 / 10 of the amount of directly transmitted image data), which effectively alleviates the pressure on cloud computing and storage capabilities and can effectively ensure data accuracy. At the same time, through multiple-pass collection and single-pass mapping, there is no need to occupy too much computing power on the vehicle side, and the efficiency is higher.

[0059] Step S202, in response to aggregated data obtained by aggregating multiple passes of data on the single-pass mapping result, obtain element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, and the element data is used to generate fused map data during association fusion.

[0060] In this embodiment, multiple-pass data, that is, data corresponding to the vehicle-side mapping results collected through multiple passes, are aggregated to facilitate analysis of the vehicle-side mapping results. Among them, the data aggregation process can be executed in the cloud or in other servers or devices, and this embodiment does not specifically limit this. Exemplarily, the aggregation process of multiple-pass data can be to align the results of each map collection in space, for example, the map elements in the single-pass map collection results are aggregated on the same map benchmark according to the data point position, so that one data point position contains the results collected corresponding to multiple passes of data, so as to facilitate multiple observations of the same map element (that is, one observation of the data corresponding to each vehicle-side mapping result). In some embodiments, data aggregation can also be performed in a time-aligned manner, and this embodiment does not specifically limit the aggregation process of multiple passes of data.

[0061] After obtaining the aggregated data, the cloud can observe the same map element multiple times for the aggregated data, that is, the aggregated data contains multiple data for the same map element, and an observation can be performed on each data. The observation results may include the location and geometric information of the map element corresponding to each data. For each map element, by integrating the observation results of multiple data, through data fusion (such as the process of determining the fusion position mentioned in the embodiments below), data comparison, etc., to obtain map elements with higher credibility and higher accuracy, and by fusing and associating these element data (according to the mapping rules in the relevant technology, the association between matching elements can be established to form map data with topological connectivity and logical relationships, and a stable and complete lane-level reality expression can be constructed), they can be fused into a map to obtain fused map data.

[0062] It can be understood that in response to the conditions or states on which the executed operations depend, when the dependent conditions or states are met, one or more operations executed can be in real time or have a set delay; unless otherwise specified, there is no restriction on the order of execution of the multiple operations executed.

[0063] Step S203: According to the matching relationship between the pre-stored existing map data and the fused map data, geometric correction is performed on the fused map data to construct a multi-source map.

[0064] In this embodiment, the existing map data can be a navigation map or a high-precision map obtained from a gallery, and these map data contain relatively accurate data geometry, topology, and attribute information. By establishing a matching relationship between the existing map data and the fused map data, and using the matching relationship to perform geometric and position corrections on the matching data, such as using high-precision map geometry data to perform registration corrections on the map geometry and supplement the broken geometry, it can effectively solve the problem that the data obtained by multi-source mapping usually only contains vehicle trajectories, photos or videos collected by forward-looking or surround-view cameras, and in scenes with poor positioning signals such as multi-layer roads, parallel roads, or tunnels, the mapping accuracy is low.

[0065] The technical solution provided in this embodiment performs single-trip mapping during the driving process of the vehicle, adopts the vehicle-side mapping method, and only needs to transmit back the vector data after the vehicle-side mapping, which greatly reduces the amount of data returned. The cloud aggregates the vehicle-side mapping results collected by the vehicle through multiple trips and observes the aggregated data multiple times to obtain more accurate map element data based on the observation results for fusion mapping, and can correct the fused mapping in combination with existing map data (such as navigation maps, high-precision map data), thereby supplementing information for some scenes with poor positioning signals, greatly optimizing the mapping effect, and improving the mapping accuracy.

[0066] Figure 3 1 is a flowchart of another method for constructing a map based on crowdsource data provided by an embodiment of the present application. Based on the above embodiment, this embodiment starts the mapping task through cloud-vehicle interaction to obtain the vehicle-side mapping result. Specifically, in addition to the above steps S201-S203, the method provided by this embodiment further divides step S201 into step S2011, and can also include the following step S301 before step S2011, such as Figure 3 As shown:

[0067] Step S301: Acquire the trajectory data transmitted by the multi-source vehicle end in real time, and match the trajectory data with the preset trajectory data used for map building to obtain a trajectory matching result.

[0068] Since the crowd-source vehicle end can be a personal vehicle, its driving trajectory can be a trajectory unrelated to mapping or a trajectory related to mapping. In order to avoid the cloud from collecting data unrelated to mapping, this embodiment obtains the trajectory data transmitted by the crowd-source vehicle end in real time before receiving the vehicle-side mapping results of the crowd-source vehicle end, and matches it with the preset trajectory data used for mapping (technical personnel in this field can adaptably determine the preset trajectory data according to the mapping needs), and the trajectory matching result is used as the basis for whether to trigger the upload of the vehicle-side mapping results.

[0069] Step S2011: When the trajectory matching result is consistent, trigger a mapping task to the multi-source vehicle end, and the mapping task is used to instruct the multi-source vehicle end to send the vehicle end mapping result to obtain the vehicle end mapping result sent by the multi-source vehicle end after multiple rounds of collection.

[0070] Exemplarily, the crowdsourcing task management system in the cloud can trigger a mapping task to the crowd-source vehicle side. The mapping task can be established by the user in the initialization stage, or it can be temporarily created when the trajectory consistency is identified. This embodiment does not specifically limit this. The mapping task is used to instruct the crowd-source vehicle side to send the vehicle-side mapping results to the cloud. The crowd-source mapping platform in the cloud pulls the vehicle-side mapping results from the crowd-source vehicle side and uses the vehicle-side mapping results to build the map.

[0071] Through the above method, using real-time trajectory data transmission and matching, the multi-source vehicle side performs vehicle-side mapping and transmission according to the mapping task, which can effectively avoid data interaction between the multi-source vehicle side and the cloud that is not related to the mapping trajectory, thereby improving mapping efficiency.

[0072] Next, the above step S202 of obtaining the element data corresponding to each map element according to the observation results of multiple observations of the same map element in the aggregated data is further described. Figure 4 As shown, step S202 may include the following steps:

[0073] Step S2021: Based on the observation results of multiple observations on the same map element in the aggregated data, element fusion processing is performed on the same map element to obtain fusion results of each map element, wherein the fusion results include fusion position information and / or fusion geometry information.

[0074] In this embodiment, the same map element may refer to the same map element at the same position. By using the observation results to perform element fusion processing on the same map element, the element position, geometric information and other data corresponding to each observation may be fused to obtain fused position information, fused geometric information, etc. Compared with a single observation data, more accurate position information and geometric information may be obtained.

[0075] In some embodiments, in order to improve the fusion accuracy of map elements, it is also possible to perform registration classification in combination with existing map data, and perform element fusion processing in combination with registration classification. Specifically, the method may also include the following steps: performing registration classification on each map element in the aggregated data according to the road topology information of the existing map data to obtain a registration classification result, and the registration classification result is used to indicate the corresponding relationship between the same map elements at the same position in the aggregated data.

[0076] For example, the same elements in multiple observation results of the same place can be matched first (for example, according to information such as the element location), and the topological information of the navigation map or high-precision map data (such as attribute information) can be obtained to align and classify the map elements of the aggregated data, so that the geometry and position of each map element are more accurate, so as to indicate the correspondence between the same map elements at the same location, so as to facilitate subsequent element fusion processing.

[0077] The above step S2021 performs feature fusion processing on the same map element according to the observation results of multiple observations on the same map element in the aggregated data, and can be carried out in the following manner: performing feature fusion processing on the same map element according to the registration classification result and the observation result.

[0078] In this embodiment, after the map elements are registered and classified, the corresponding relationship between the same map elements at the same position is clearer, and the map elements have a topological relationship with high precision, so the data with low precision in the observation results can be eliminated (for example, the topological relationship of the same map element in a certain observation result in the aggregated data is inconsistent with the map element in other observation results, or for a certain map element in the aggregated data, the position difference or geometric difference with the corresponding map elements in other observations is too large, the map element corresponding to the observation can be eliminated, and the observation data after elimination is subjected to element fusion processing.

[0079] Through this process, the topological information and multiple observation results in the existing map data can be effectively utilized to accurately fuse map elements, thereby improving the accuracy of map data.

[0080] Step S2022: Determine the position information and / or geometric information corresponding to each map element according to the fused position information and / or the fused geometric information, and obtain element data of each map element according to the position information and / or geometric information.

[0081] In this embodiment, for each map element, the fused position information and fused geometric information corresponding to multiple observation results are used to determine the position information and geometric information of the map element. Compared with a single observation, this is more reliable and provides data support for improving mapping accuracy.

[0082] Through the above technical solution, for each map element, multiple observation results of the same map element are fused and processed to obtain a more reliable and accurate result.

[0083] In some embodiments, in addition to the data fusion method, a credibility calculation method can also be used to select the map element with the highest credibility from the observation results for the same map element as the matching element, wherein the credibility calculation can be obtained based on the deviation between each observation result and other observation results, wherein the larger the deviation, the lower the credibility.

[0084] Figure 5 This is a flowchart of another method for constructing a map based on crowdsource data provided by an embodiment of the present application. On the basis of the above embodiment, this embodiment uses the constructed crowdsource map to update the high-precision map. Specifically, in addition to the above steps S201-S203, after the crowdsource map is constructed in step S203, the method provided by this embodiment may also include the following steps S501-S503.

[0085] Step S501: for each map element in the crowd-source map, obtain at least one of a consistency recognition result between the map element and a corresponding map element in the aggregated data, and a positioning accuracy corresponding to the map element.

[0086] It can be understood that each map element in the crowd-source map is map element data after the element data of each map element is matched and corrected with the existing map data.

[0087] In one example, the confidence evaluation of map elements is achieved by obtaining the consistency recognition results of map elements, and the map elements in the multi-source map are identified with the corresponding map elements in the aggregated data for consistency, that is, the convergence process of multiple data results (the process of making multiple observation results consistent and stable through data fusion), and the accuracy of the map elements in the generated multi-source map can be inferred. Among them, the consistency recognition process refers to the consistency between the location information of the map elements of the multi-source map and the corresponding map elements in the aggregated data (if the distance is less than the distance threshold, the elements are considered to be consistent, and the distance threshold can be determined based on actual applications or prior data).

[0088] In another example, the positioning accuracy corresponding to the map element can also be used to evaluate the confidence of the map element. For example, if the map element is the map element with the highest accuracy selected after multiple data fusions, the positioning accuracy can be determined based on the GPS signal carried by the single data collection (data source) to which the map element belongs. The higher the GPS signal strength, the higher the positioning accuracy.

[0089] In some instances, the consistency recognition result and positioning accuracy of map elements can be used simultaneously. For example, corresponding weights can be assigned to these two indicators respectively, and the confidence of the map element can be jointly calculated, such as by using the following formula: y=ax1+bx2, where y represents the confidence of a map element, x1 represents the consistency recognition result of the map element, x2 represents the positioning accuracy of the map element, and a and b are the weights assigned to the consistency recognition result and positioning accuracy according to prior data.

[0090] Step S502: Calculate the confidence of each map element in the crowd-source map according to the consistency recognition result of each map element and at least one of the positioning accuracy corresponding to the map element.

[0091] For the calculation process of the consistency recognition result of the map element, for example, the aggregated data is obtained for the mapping results of 8 trips of the vehicle side, and for each map element in the crowd source map, the consistency recognition result between the map element and the corresponding element of the mapping results of the 8 trips of the vehicle side in the aggregated data is calculated respectively. For example, the element data of the map element is consistent with the element data of the map element in 7 trips of data (the distance is less than the distance threshold), but inconsistent with the element data in 1 trip of data, then the consistency recognition result can be 7 / 8, wherein the consistency recognition result corresponds to the confidence (can be equal or proportional, and this embodiment is not particularly limited to this), and the higher the consistency in the consistency recognition result, the higher the confidence. Correspondingly, the positioning accuracy of the map element, for example, the GPS signal is L1 C / A code, and its positioning accuracy is 5 to 10 meters. For the convenience of calculation, the index value of quantifying the positioning accuracy is 0.5, and the GPS signal is real-time kinematic (RTK), and its positioning accuracy is centimeter level, within 1cm to 2cm, and its index value can be quantified as 1, and so on.

[0092] In some embodiments, when one of the indicators is used to calculate the confidence, it can also be combined with a pre-set corresponding confidence mapping table, and the confidence calculated based on the consistency recognition result is taken as an example for explanation. For example, the confidence can be divided into multiple levels. The consistency recognition result is in interval a, corresponding to confidence 1, interval b, corresponding to confidence 2, the consistency recognition result is in interval c, corresponding to confidence 3, ..., and so on. The present application does not specifically limit the specific calculation method of the confidence.

[0093] Step S503: perform data differentiation on the map elements with a confidence level higher than the confidence level threshold and the map elements in the latest version of the high-precision map data, and update the high-precision map data according to the differentiation result.

[0094] It should be noted that those skilled in the art can adaptively determine the confidence threshold in combination with practical applications or prior data. Map elements with a confidence level higher than the confidence threshold indicate that they have a higher accuracy and can be used to update high-precision map data. The data differencing process, that is, identifying the differences between map elements above the confidence threshold (which can be identified item by item) and the map elements in the current latest version of high-precision map data, such as changes in road structure, new buildings, updates to traffic signs, etc., to obtain the changed map elements, and use the changed map elements to update the high-precision map data.

[0095] In some embodiments, low to medium confidence data (i.e., data below a confidence threshold) can be manually confirmed in the high-precision map production line. Elements manually confirmed as correct for mapping can be transferred to the high-confidence element process, and results confirmed as errors can be included in the sample library and can be used to optimize the algorithm later.

[0096] In some embodiments, after the confidence level of map elements is evaluated, in addition to using these map element data to update high-precision map data, mapping and publishing processes can also be performed based on these map data.

[0097] It should be noted that the high-precision map data in this embodiment and the above-mentioned existing map data can be the same map data or different map data, and this embodiment does not specifically limit this.

[0098] This embodiment converges the results of multiple data passes to determine the confidence of the mapping results of each element. The data with high confidence is input into the high-precision map production line for differentiation and automatic update of the high-precision map data, which can effectively improve the update efficiency and accuracy of the high-precision map.

[0099] Figure 6 : is a flowchart of another method for constructing a map based on crowd-source data provided by an embodiment of the present application, which is applied to a crowd-source vehicle side. The method includes the following steps S601-S603:

[0100] Step S601: collect map perception data during driving, and perform vectorization processing on the map perception data to obtain a single-trip mapping result.

[0101] In this embodiment, the crowd-source vehicle end can be an operating vehicle or a personal vehicle. Instead of directly transmitting the collected image data to the cloud for mapping in related technologies, the crowd-source vehicle end can collect road images (or other road data such as road point cloud data) when receiving a mapping task triggered by the cloud, and use the built-in perception algorithm to recognize the image to obtain map perception data, and obtain a single-trip mapping result by vectorizing the map perception data.

[0102] Among them, vectorization processing is to convert map perception data into vector form, that is, to convert the original image data into a data set of geometric figures (such as line segments, curves, polygons, etc.) and attributes (such as position, direction, speed, etc.), and the data at different time points can be aligned and fused to form a single-trip mapping result, which can include detailed information on the vehicle's driving path and its surrounding map environment.

[0103] Step S602: Based on the multi-pass collection process of the above steps, a vehicle-side mapping result after multi-pass collection is obtained.

[0104] In this embodiment, the multi-trip collection is the multi-trip collection for the same road section. The multi-trip collection process can be the multi-trip data collected by the same vehicle in the multi-source vehicle terminal, or the multi-trip data collected by multiple vehicles together, which is not particularly limited in this embodiment.

[0105] Step S603: Send the vehicle-side mapping results collected through multiple rounds to the cloud, so that the cloud builds a multi-source map based on the vehicle-side mapping results.

[0106] It should be noted that the specific construction process of the crowd-source map can refer to the above embodiment, and other related instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the embodiment of the method of this application, and no further details will be given here.

[0107] Next, the map perception data is vectorized in the above step S601 to obtain a single-pass mapping result, which is further described as follows: Figure 7a As shown, the following steps S6011-S6013 are included:

[0108] Step S6011: Map the map perception data to an absolute coordinate system according to the posture information and calibration parameters of the multi-source vehicle end to obtain the perception data after coordinate conversion.

[0109] The image perception data collected by the crowd-source vehicle side is the perception data corresponding to the image, which is data based on the image coordinate system. In order to facilitate the processing of the data in the cloud, the data of the image coordinate system is converted into an absolute coordinate system (ie, geographic coordinate system). It can be understood that the posture information can be obtained from the vehicle's sensors (such as GPS, IMU, etc.) during driving, that is, inertial navigation positioning information, including position and direction, etc. The calibration parameters are fixed parameters used to correct sensor data to ensure its accuracy. These parameters usually include the sensor's intrinsic parameters (such as focal length, principal point position) and extrinsic parameters (rotation and translation of the sensor relative to the vehicle coordinate system).

[0110] Exemplarily, the above process of mapping the map perception data to the absolute coordinate system can be: first, using the vehicle's camera extrinsics, the map perception data is converted from the camera's local coordinate system (i.e., the image coordinate system) to the vehicle coordinate system; then, using the vehicle's posture information, the data is converted from the vehicle coordinate system to the absolute coordinate system to achieve the mapping of the perception data to the absolute coordinate system.

[0111] Step S6012: motion compensate the perception data according to a preset motion distortion equation, and track the map elements of the previous and next frames on the motion compensated perception data to obtain tracked multi-frame perception data.

[0112] When the vehicle is moving, the data collected by the sensor may be distorted due to the movement. In this embodiment, motion compensation is performed on the perception data through a preset motion distortion equation (technical personnel in this field can determine the motion distortion equation based on practical applications or prior data) to correct these distortions.

[0113] After motion compensation of the perception data, the same map elements (such as feature points or objects in the perception data) between consecutive frames are identified and tracked to achieve tracking of the same features of multi-frame data to ensure the continuity and consistency of multi-frame data.

[0114] Step S6013: vectorize the multi-frame perception results to obtain a single-pass mapping result.

[0115] In this embodiment, multiple frames of perception data are converted into vector format to achieve vector reconstruction, such as conversion into geometric figures (such as line segments, curves, polygons, etc.) and attributes (such as position, direction, speed, etc.), and these vectorized data are integrated into a complete map, which contains detailed information about the vehicle's driving path and its surrounding environment, that is, a single-trip mapping result.

[0116] Through the above technical solution, by utilizing methods such as coordinate system conversion, motion compensation, and feature tracking, the perception data collected during vehicle driving can be effectively converted into a single-trip mapping result with higher accuracy, providing high-precision data support for multi-source mapping in the cloud.

[0117] Next, the process of tracking map elements of the previous and next frames on the motion-compensated perception data in the above step S6012 is further introduced: according to a preset feature extraction algorithm, map element features in the motion-compensated perception data are extracted; the map element features are associated between the previous and next frames to obtain an association result of the map element features; based on the association result, map element tracking is performed on the motion-compensated perception data with respect to the previous and next frames.

[0118] Exemplarily, the preset feature extraction algorithm can extract map element features (such as road boundaries, lane lines, building outlines, traffic signs, etc.) by using an edge detection algorithm to extract key features, or use a scale-invariant feature transform (SIFT) algorithm to detect map element features in an image. In the feature association process, feature association can be achieved by feature matching. Specifically, for each previous and next frame, the most similar feature can be found by comparing the distance between map element features (such as Euclidean distance), or a fast nearest neighbor algorithm can be used to match and associate features. In some embodiments, an optical flow method can also be used to track the movement of feature points in continuous frames to achieve feature association. This embodiment does not specifically limit the specific association process of map element features. After feature association, a filter (such as a Kalman filter or a particle filter) can be used to estimate and predict the position and state of the feature in the time series, so as to track the position and corresponding trajectory of each map element in continuous frames according to the result of feature association.

[0119] Furthermore, the above-mentioned step S6013 vectorizes the multi-frame perception results to obtain a single-pass mapping result, which can be done in the following manner: converting the format of the map elements in the multi-frame perception results into a vector format to obtain vectorized map elements; constructing a topological structure based on the adjacency relationship between the map elements in the multi-frame perception results; and constructing a single-pass map based on the vectorized map elements and the topological structure to obtain the single-pass mapping result.

[0120] Specifically, the image format data or point cloud format data corresponding to the multi-frame perception results are converted into geometric figures such as points, lines, and surfaces to represent map elements. For example, road boundaries are represented as a series of connected line segments, and buildings are represented as polygons. And by analyzing the spatial relationship between map elements in the multi-frame perception results, their adjacency relationship is identified. The adjacency relationship can be determined by calculating the distance, intersection, and overlap between map elements, so as to construct a topological structure based on the adjacency relationship. For example, the topological structure can be constructed using graph theory methods, with map elements as nodes and adjacency relationships as edges, so as to understand the connection and relative position between map elements and support applications such as path planning and navigation. Based on this, by integrating vectorized elements and topological structures, combining vectorized map elements with the constructed topological structure, a complete map representation can be formed, thereby constructing a single-trip map constructed by the data collected by the multi-source vehicle during the current driving process.

[0121] The following is a further example of the above process. Figure 7b As shown in the figure, the single-trip mapping process of the crowd-source vehicle side can include the following process:

[0122] a. The vehicle uploads trajectory data in real time during driving, and matches the trajectory with the cloud task management system. The matching result is used as the basis for triggering the upload of vehicle-side mapping results.

[0123] b. During driving, the image results are saved in real time, and the image perception module is triggered in the background to output single-frame image perception results using a deep learning algorithm.

[0124] c. The image perception module results are based on the image coordinate system. Combined with the fused inertial navigation and calibration result parameters, the image coordinates are converted to the absolute coordinate system in the pose fusion link.

[0125] d. Motion compensation and tracking prediction: The perception results are optimized by establishing motion distortion compensation equations, and the same features of multi-frame data are tracked through algorithms to reduce the computing power requirements of the terminal device.

[0126] For the above steps 2-4, the perception result of the single-frame image output in step 2 is used. At this time, the perception result is still in the image coordinate system. After step 3, the result is changed to the absolute coordinate WGS84 or 02 coordinate. In step 4, the image distortion caused by motion is eliminated, and the same perception result is associated with the images of the previous and next frames to obtain the timing result. For example, if a marking line is recognized in the first frame, the corresponding marking line can be found in the subsequent second frame, third frame, etc. through step 4, so as to form a continuous marking line later.

[0127] e. Based on the multi-frame perception results obtained in the previous stage, the multi-frame perception results are vectorized to obtain the single-trip mapping results on the vehicle side.

[0128] The above technical solution of this embodiment reduces the dependence of high-precision mapping on point clouds and high-precision positioning equipment through the crowd-source mapping process of visual mapping technology innovation, and realizes high-precision map construction and updating through multiple observations combined with the high-precision map data ecosystem.

[0129] Figure 8 is a schematic diagram of the interaction between the multi-source vehicle end and the cloud in an embodiment of the present application, such as Figure 8 As shown, the following process is included:

[0130] Step S801: The multi-source vehicle transmits trajectory data to the cloud in real time during driving;

[0131] Step S802: The cloud triggers a mapping task according to the matching result of the trajectory data;

[0132] Step S803: The multi-source vehicle collects map perception data during driving, and performs vectorization processing on the map perception data to obtain a single-trip mapping result;

[0133] Step S804: The multi-source vehicle-side transmits the vehicle-side mapping results collected through multiple rounds to the cloud;

[0134] Step S805: The cloud obtains the vehicle-side mapping results sent by the multi-source vehicle-side after multiple rounds of collection, and aggregates the multiple rounds of data to obtain aggregated data;

[0135] Step S806: The cloud obtains element data corresponding to each map element according to the observation results of multiple observations of the same map element in the aggregated data, and associates and fuses these element data to obtain fused map data;

[0136] Step S807: The cloud performs geometric correction on the fused map data according to the matching relationship between the existing map data and the fused map data to construct a multi-source map.

[0137] Further, in order to facilitate the understanding of the embodiments of the present application, Fig. 9 As shown, it includes the mapping stage S901, the mapping data pulling stage S902, the crowd-source map construction stage S903 and the map update stage S904; wherein in the mapping stage S901, the vehicle-side backend of the crowd-source vehicle uploads the trajectory data to the task management platform on the cloud in real time, and the task management platform matches the trajectory data. If the trajectory matches, the crowd-sourced task management system of the cloud-based task management platform in the mapping data pulling stage S902 pulls the vehicle-side mapping results from the vehicle-side backend, and the vehicle-side backend sends the vehicle-side mapping results to the crowd-sourced task management system. Then, in the crowd-source map construction stage S903, the cloud-based crowd-source mapping platform executes the following mapping process according to the vehicle-side mapping results:

[0138] 1. After receiving the task, in the data aggregation phase, obtain the vehicle-side mapping results, collect data on the same road section, and trigger the next phase after meeting the number of trips required.

[0139] 2. Perform registration and classification on the aggregated data, and match the same elements in multiple observations of the same place. At the same time, through the established matching relationship, the topological information and attribute information of the navigation map or high-precision map data can be obtained to solve the problem of incorrect mapping results due to insufficient trajectory accuracy in scenarios with poor positioning signals such as multi-layer roads, parallel roads, or in tunnels.

[0140] 3. Based on the registration and classification results, multiple observations of the same object (i.e. map element) are fused to obtain more reliable and accurate results. At this time, each object exists independently. For example, the marking only has geometry and type information, and does not form a lane with other markings or boundaries.

[0141] 4. The fused data are associated with each other according to the rules to form map data with topological connectivity and logical relationships, and to build a stable and complete lane-level reality expression.

[0142] 5. The map registration and fusion mapping module establishes a matching relationship between the existing navigation / high-precision map data and the aggregated map data, uses the high-precision map geometry data to register and correct the mapping geometry, and supplements the broken geometry to increase the accuracy and correctness of the mapping data.

[0143] 6. Confidence evaluation: by converging the results of multiple data passes, the confidence of the mapping results of each element is judged. For data with high execution degree, it is directly input into the high-precision map production line for differential and automatic update of high-precision map data in the map update stage S904. For medium and low confidence data, it will be manually confirmed in the high-precision map production line, and the correct mapping elements will be transferred to the high-confidence element process. The results confirmed as wrong will be included in the sample library and can be used to optimize the algorithm later.

[0144] The above technical solutions in the embodiments of the present application replace the professional high-precision data collection vehicle solution. The crowd-source mapping method of the present application can greatly reduce the cost of data collection. The cost of a single vehicle is reduced from 600,000 yuan to the level of 1,000 yuan. In addition, through the means of multiple observations, the accuracy and correctness of mapping are supplemented to meet the needs of high-precision maps. Compared with crowd-source mapping in related technologies, which uses source data such as vehicle-transmitted videos to build maps in the cloud, the present application implements single-trip mapping based on low-computing chips on the vehicle side, greatly reducing the cost of data transmission and storage. The data volume is only 1 / 100 of the video return solution; in addition, the present application solves some crowd-source mapping problems in special scenarios by combining navigation or high-precision map data, such as the road geometry and connectivity of the navigation map. In scenarios with poor positioning signals such as multi-layer roads and tunnels, trajectory matching errors lead to matching the wrong road when the vehicle-side mapping data is aggregated in the cloud.

[0145] Fig.101 is a schematic diagram of the structure of a map construction device based on crowdsource data provided in an embodiment of the present application, which is applied to the cloud. The device 100 includes a first acquisition module 101, a second acquisition module 102 and a first mapping module 103, wherein the first acquisition module 101 is configured to acquire a vehicle-side mapping result sent by a crowdsource vehicle-side after multiple collections according to a pre-triggered mapping task, wherein the vehicle-side mapping result is a single-pass mapping result obtained after the crowdsource vehicle-side collects map perception data and vectorizes the map perception data; the second acquisition module 102 is configured to respond to aggregated data obtained by performing multiple data aggregation on the single-pass mapping result, and acquire element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, wherein the element data is used to generate fused map data during association fusion; the first mapping module 103 is configured to perform geometric correction on the fused map data according to a matching relationship between pre-stored existing map data and the fused map data to construct a crowdsource map.

[0146] In one embodiment, the device also includes: a trajectory acquisition module, which is configured to acquire the trajectory data transmitted by the multi-source vehicle end in real time, and match the trajectory data with the preset trajectory data used for mapping to obtain a trajectory matching result; the first acquisition module 101 is specifically configured to: when the trajectory matching result is consistent, trigger a mapping task to the multi-source vehicle end, and the mapping task is used to instruct the multi-source vehicle end to send the vehicle-end mapping result, so as to obtain the vehicle-end mapping result sent by the multi-source vehicle end after multiple rounds of collection.

[0147] In one embodiment, the second acquisition module 102 includes: a fusion unit, which is configured to perform element fusion processing on the same map element based on the observation results of multiple observations of the same map element in the aggregated data, to obtain a fusion result of each map element, wherein the fusion result includes fusion position information and / or fusion geometric information; a determination unit, which is configured to determine the position information and / or geometric information corresponding to each map element based on the fusion position information and / or the fusion geometric information, and obtain the element data of each map element based on the position information and / or geometric information.

[0148] In one embodiment, the device also includes: a registration and classification module, which is configured to perform registration and classification on each map element in the aggregated data based on the road topology information of the existing map data to obtain a registration and classification result, and the registration and classification result is used to indicate the correspondence between the same map elements at the same position in the aggregated data; the fusion unit is specifically configured to perform element fusion processing on the same map element based on the registration and classification result and the observation result.

[0149] In one embodiment, the device also includes: an identification module, which is configured to obtain, for each map element in the crowd-source map, a consistency identification result between the map element and the corresponding map element in the aggregated data, and at least one of the positioning accuracies corresponding to the map element; a confidence calculation module, which is configured to calculate the confidence of each map element in the crowd-source map according to the consistency identification result of each map element and at least one of the positioning accuracies corresponding to the map element; a differential update module, which is configured to perform data differentiation between the map elements whose confidence is higher than the confidence threshold and the corresponding map elements in the latest version of the high-precision map data, and update the corresponding map elements in the high-precision map data according to the differential result.

[0150] Fig.11 is a schematic diagram of a map construction device based on crowd-source data provided in an embodiment of the present application, which is applied to crowd-source vehicle terminals, such as Fig.11 As shown, the device 110 includes a second mapping module 111, a third acquisition module 112 and a transmission module 113, wherein the second mapping module 111 is configured to collect map perception data during driving, and vectorize the map perception data to obtain a single-pass mapping result; the third acquisition module 112 is configured to obtain a vehicle-side mapping result after multiple-pass collection based on a multiple-pass collection process of the second mapping module; the transmission module 113 is configured to send the vehicle-side mapping result after multiple-pass collection to the cloud, so that the cloud can build a multi-source map based on the vehicle-side mapping result.

[0151] In one embodiment, the second mapping module 111 includes: a coordinate conversion unit, which is configured to map the map perception data to an absolute coordinate system according to the posture information and calibration parameters of the multi-source vehicle end to obtain the perception data after coordinate conversion; a compensation tracking unit, which is configured to perform motion compensation on the perception data according to a preset motion distortion equation, and perform map element tracking processing on the previous and next frames on the motion-compensated perception data to obtain tracked multi-frame perception data; a vectorization processing unit, which is configured to perform vectorization processing on the multi-frame perception results to obtain a single-pass mapping result.

[0152] In one embodiment, the compensation tracking unit is specifically configured to: extract map element features from motion-compensated perception data according to a preset feature extraction algorithm; associate the map element features between previous and next frames to obtain association results of the map element features; and perform map element tracking processing on the motion-compensated perception data with respect to previous and next frames according to the association results.

[0153] In one embodiment, the vectorization processing unit is specifically configured to: convert the format of the map elements in the multi-frame perception results into a vector format to obtain vectorized map elements; construct a topological structure based on the adjacency relationship between the map elements in the multi-frame perception results; and construct a single-pass map based on the vectorized map elements and the topological structure to obtain the single-pass mapping result.

[0154] The present application also provides an electronic device, which may be a cloud server (corresponding to the cloud in the above embodiment). Fig.12 As shown, the cloud server includes: a processor 121, and a memory 122 connected to the processor 121 in communication; the memory 122 stores computer-executable instructions; the processor 121 executes the computer-executable instructions stored in the memory 122 to implement Figure 2-Figure 5 The corresponding embodiment provides a map construction method based on crowd-source data, wherein the memory 122 and the processor 121 are connected via a bus 123 .

[0155] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0156] The present application also provides another electronic device, which may be a vehicle (corresponding to the multi-source vehicle end in the above embodiment). Fig.13 As shown, the vehicle includes: a processor 131, and a memory 132 in communication with the processor 131; the memory 132 stores computer-executable instructions; the processor 131 executes the computer-executable instructions stored in the memory 122 to implement Figure 6-7b The corresponding embodiment provides a map construction method based on crowd-source data, wherein the memory 132 and the processor 131 are connected via a bus 133 .

[0157] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0158] The present application also provides a system for constructing a map based on crowdsource data. Fig.14 As shown, the system 140 includes a cloud 141 and a multi-source vehicle end 142 electrically connected to the cloud 141; the cloud 141 is used to execute the above Figure 2-Figure 5 The corresponding embodiment provides a method for constructing a map based on crowdsource data; the crowdsource vehicle terminal 142 is used to perform the above Figure 6-7b The corresponding embodiment provides a map construction method based on crowd-source data.

[0159] The present application embodiment provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the computer-executable instructions are used to implement Figure 2-Figure 5 The corresponding embodiment provides a map construction method based on crowd-source data.

[0160] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0161] The present application embodiment provides another computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, the computer-executable instructions are used to implement Figure 6-7b The corresponding embodiment provides a map construction method based on crowd-source data.

[0162] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0163] Among them, the computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0164] The present application provides a computer program product, wherein the computer program product includes a computer program, which is executed by a processor to implement Figure 2-Figure 5 The corresponding embodiment provides a map construction method based on crowd-source data.

[0165] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0166] The present application embodiment provides another computer program product, wherein the computer program product includes a computer program, which is executed by a processor to implement Figure 6-7b The corresponding embodiment provides a map construction method based on crowd-source data.

[0167] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0168] The embodiment of the present application also provides a chip, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory to execute the map construction method based on crowd-source data.

[0169] The relevant instructions can be understood by referring to the relevant descriptions and effects corresponding to the steps in the method embodiments of the present application, and no further elaboration is made here.

[0170] Fig.15 It is a block diagram of a terminal device shown in an exemplary embodiment of the present application. The terminal device 800 can be a mobile phone, a computer, a digital broadcast terminal, a message transceiver device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0171] The terminal device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0172] The processing component 802 generally controls the overall operation of the terminal device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0173] The memory 804 is configured to store various types of data to support operations on the terminal device 800. Examples of such data include instructions for any application or method operating on the terminal device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0174] The power supply component 806 provides power to various components of the terminal device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the terminal device 800.

[0175] The multimedia component 808 includes a screen that provides an output interface between the terminal device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the terminal device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0176] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the terminal device 800 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0177] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0178] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the terminal device 800. For example, the sensor assembly 814 can detect the open / closed state of the terminal device 800, the relative positioning of the components, such as the display and keypad of the terminal device 800, and the sensor assembly 814 can also detect the position change of the terminal device 800 or a component of the terminal device 800, the presence or absence of contact between the user and the terminal device 800, the orientation or acceleration / deceleration of the terminal device 800, and the temperature change of the terminal device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0179] The communication component 816 is configured to facilitate wired or wireless communication between the terminal device 800 and other devices. The terminal device 800 can access a wireless network based on a communication standard, such as WiFi, 3G, 4G, 5G or other standard communication networks, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0180] In an exemplary embodiment, the terminal device 800 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the method provided in any of the above-mentioned method embodiments of the present application.

[0181] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by a processor 820 of a terminal device 800 to complete the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0182] An embodiment of the present application also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the terminal device, the terminal device 800 can execute the method provided by any one of the above-mentioned method embodiments of the present application.

[0183] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0184] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0185] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for constructing a map based on crowdsource data, characterized in that: Applied in the cloud, including: According to the pre-triggered mapping task, a vehicle-side mapping result sent by the crowd-source vehicle-side after multiple rounds of collection is obtained, wherein the vehicle-side mapping result is a single-pass mapping result obtained after the crowd-source vehicle-side collects map perception data and performs vectorization processing on the map perception data; In response to aggregated data obtained by aggregating multiple passes of data on the single pass mapping result, obtaining element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, wherein the element data is used to generate fused map data during association fusion; According to the matching relationship between the pre-stored existing map data and the fused map data, the fused map data is geometrically corrected to construct a multi-source map.

2. The method according to claim 1, characterized in that: Also includes: Acquire the trajectory data transmitted by the multi-source vehicle end in real time, and match the trajectory data with the preset trajectory data used for map building to obtain a trajectory matching result; The obtaining of the vehicle-side mapping results sent by the multi-source vehicle-side after multiple rounds of collection includes: When the trajectory matching result is consistent, a mapping task is triggered to the multi-source vehicle end, and the mapping task is used to instruct the multi-source vehicle end to send a vehicle-end mapping result to obtain the vehicle-end mapping result sent by the multi-source vehicle end after multiple rounds of collection.

3. The method according to claim 1 or 2, characterized in that: The acquiring the element data corresponding to each map element according to the observation results of multiple observations of the same map element in the aggregated data includes: According to the observation results of multiple observations of the same map element in the aggregated data, element fusion processing is performed on the same map element to obtain fusion results of each map element, wherein the fusion results include fusion position information and / or fusion geometric information; The position information and / or geometric information corresponding to each map element is determined according to the fused position information and / or the fused geometric information, and the element data of each map element is obtained according to the position information and / or the geometric information.

4. The method according to claim 3, characterized in that: Also includes: According to the road topology information of the existing map data, register and classify each map element in the aggregated data to obtain a registration and classification result, wherein the registration and classification result is used to indicate the corresponding relationship between the same map elements at the same position in the aggregated data; The performing element fusion processing on the same map element according to the observation results of multiple observations on the same map element in the aggregated data includes: According to the registration classification result and the observation result, element fusion processing is performed on the same map element.

5. The method according to any one of claims 1 to 4, characterized in that: After building the crowdsource map, it also includes: For each map element in the crowd-source map, obtaining at least one of a consistency recognition result between the map element and a corresponding map element in the aggregated data, and a positioning accuracy corresponding to the map element; Calculating the confidence of each map element in the crowd-source map according to the consistency recognition result of each map element and at least one of the positioning accuracy corresponding to the map element; The map elements with confidence levels higher than the confidence level threshold are differentiated from the corresponding map elements in the latest version of the high-precision map data, and the corresponding map elements in the high-precision map data are updated according to the differentiation results.

6. A method for constructing a map based on crowdsource data, characterized in that: Applied to multi-source vehicle terminals, including: Collecting map perception data during driving, and performing vectorization processing on the map perception data to obtain a single-trip mapping result; Based on the multi-pass collection process in the above steps, the vehicle-side mapping result after multiple-pass collection is obtained; The vehicle-side mapping results collected through multiple rounds are sent to the cloud, so that the cloud can build a multi-source map based on the vehicle-side mapping results.

7. The method according to claim 6, characterized in that The vectorization processing of the map perception data to obtain a single-pass mapping result includes: According to the posture information and calibration parameters of the multi-source vehicle end, the map perception data is mapped to an absolute coordinate system to obtain the perception data after coordinate transformation; Performing motion compensation on the perception data according to a preset motion distortion equation, and performing map element tracking processing on the previous and next frames on the motion-compensated perception data to obtain a tracked multi-frame perception result; Vectorization is performed on the multi-frame perception results to obtain a single-pass mapping result.

8. The method according to claim 7, characterized in that The processing of tracking map elements of previous and next frames on the motion-compensated perception data includes: Extracting map element features from the motion-compensated perception data according to a preset feature extraction algorithm; Associating the map element features between the previous and next frames to obtain a correlation result of the map element features; According to the association result, map element tracking processing is performed on the motion compensated perception data with respect to the previous and next frames.

9. The method according to claim 7 or 8, characterized in that: The vectorization processing of the multi-frame perception results to obtain a single-pass mapping result includes: Converting the format of the map elements in the multi-frame perception results into a vector format to obtain vectorized map elements; Constructing a topological structure according to the adjacency relationship between map elements in the multi-frame perception results; A single-pass map is constructed according to the vectorized map elements and the topological structure to obtain the single-pass mapping result.

10. A map construction device based on crowd-source data, characterized in that: Applied in the cloud, including: A first acquisition module is configured to acquire a vehicle-side mapping result sent by a multi-source vehicle-side after multiple-pass collection according to a pre-triggered mapping task, wherein the vehicle-side mapping result is a single-pass mapping result obtained after the multi-source vehicle-side collects map perception data and performs vectorization processing on the map perception data; A second acquisition module is configured to obtain, in response to aggregated data obtained by aggregating multiple passes of data on the single pass mapping result, element data corresponding to each map element according to observation results of multiple observations of the same map element in the aggregated data, wherein the element data is used to generate fused map data during association fusion; The first mapping module is configured to perform geometric correction on the fused map data according to a matching relationship between the pre-stored existing map data and the fused map data, so as to construct a multi-source map.

11. A map construction device based on crowd-source data, characterized in that: Applied to multi-source vehicle terminals, including: A second mapping module is configured to collect map perception data during driving and perform vectorization processing on the map perception data to obtain a single-trip mapping result; A third acquisition module, which is configured to obtain a vehicle-side mapping result after multiple acquisitions based on the multiple acquisitions of the second mapping module; The transmission module is configured to send the vehicle-side mapping results collected through multiple rounds to the cloud, so that the cloud can build a multi-source map based on the vehicle-side mapping results.

12. An electronic device / computer-readable storage medium / computer program product, characterized in that: include: Memory and processor; The memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, so that the electronic device executes the method for constructing a map based on crowd-source data according to any one of claims 1 to 9; and / or, The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method for constructing a map based on crowd-source data as described in any one of claims 1 to 9; and / or, The computer program product comprises a computer program, which, when executed by a processor, implements the method for constructing a map based on crowd-source data as described in any one of claims 1 to 9.

13. A map construction system based on crowdsource data, characterized in that: It includes a cloud, and a multi-source vehicle terminal electrically connected to the cloud; The cloud is used to execute the method for constructing a map based on crowd-source data as described in any one of claims 1 to 5; The crowd-source vehicle terminal is used to execute the map construction method based on crowd-source data as described in any one of claims 6-9.

Images