Map updating method and device, computer device and storage medium

Abstract

This application relates to a map updating method, apparatus, computer equipment, storage medium, and computer program product. The method includes: acquiring a structured data source within a target area; determining a feature vector corresponding to the structured data source based on preset data characteristics; determining the accuracy probability of the structured data source based on the feature vector and a target big data model; generating a current local map describing real-world road information based on the target area and the structured data source when the accuracy probability is higher than a preset accuracy probability threshold; determining the difference information between the current local map and the global historical map; and updating the map element information of the global historical map based on the difference information. Using this method, elements and reliable reliability information can be provided for autonomous driving maps, enabling large-scale and rapid updates of autonomous driving maps.

Description

Technical Field

[0001] This application relates to the field of map data updating technology, and in particular to a map updating method, apparatus, computer equipment, storage medium and computer program product. Background Technology

[0002] With the development of autonomous driving technology, autonomous driving maps have emerged. Autonomous driving maps are a necessary condition for achieving highly automated driving or even driverless driving. They play a key role in the perception, prediction, planning, and decision-making modules of autonomous vehicles, enabling them to build "long-term memory" and achieve beyond-visual perception.

[0003] Autonomous driving maps need to meet the requirements of high freshness, large scope, and low cost of updates. Currently, the mainstream method for updating autonomous driving maps in the industry is to obtain road perception information through crowdsourcing, and then use AI (Artificial Intelligence) recognition and extraction technology to identify and extract the perception data obtained through crowdsourcing, calculate change information, and thus obtain map change elements and map change information that can be used for updating autonomous driving maps.

[0004] However, currently, after obtaining map change information that can be used for updates, one method for autonomous driving maps is to directly update them. This method results in low reliability of the updates. To ensure that the updated map meets the quality standards, manual confirmation and operation processes are required. This method leads to low update efficiency and cannot meet the map freshness requirements of autonomous driving. Summary of the Invention

[0005] Therefore, it is necessary to provide a map updating method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the reliability of autonomous driving map updates in response to the above-mentioned technical problems.

[0006] Firstly, this application provides a map updating method. The method includes:

[0007] Obtain structured data sources within the target area;

[0008] Based on preset data characteristics, determine the feature vector corresponding to the structured data acquisition data source;

[0009] Based on the feature vectors corresponding to the structured data source and the target big data model, determine the accuracy probability of the structured data source;

[0010] If the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source.

[0011] The differences between the current local map and the global historical map are determined, and the map feature information of the global historical map is updated based on the differences.

[0012] In one embodiment, determining the feature vector corresponding to the structured acquisition data source includes:

[0013] Based on the structured data source, determine the map feature data of the structured data source for the preset data features;

[0014] Based on the map feature data and the target coding strategy, determine the coded data corresponding to the map feature data;

[0015] Based on the encoded data, the feature vector corresponding to the structured acquisition data source is determined.

[0016] In one embodiment, before determining the accuracy probability of the structured data source based on the feature vector corresponding to the structured data source and the target big data model, the method further includes:

[0017] Obtain a structured sample dataset, which includes multiple sample map data. The sample map data contains map feature information, data collection information, and feature true / false labels corresponding to the sample map data.

[0018] Determine the feature vector corresponding to each of the collected sample map data in the structured collected sample dataset, and construct the target sample dataset based on the feature vector corresponding to each of the collected sample map data and the true / false labels of the features;

[0019] The target sample dataset is input into the initial big data model for model training to obtain the target big data model.

[0020] In one embodiment, when the accuracy probability is higher than a preset accuracy probability threshold, generating a current local map describing real-world road information based on the target area and the structured data source includes:

[0021] If the accuracy probability is higher than a preset accuracy probability threshold, the structured data acquisition data source will be used as the input data source for the current local map data.

[0022] The input data source is subjected to multi-source data fusion processing to obtain the target map data;

[0023] Based on the target map data and the target area, a current local map describing real-world road information is generated.

[0024] In one embodiment, determining the difference information between the current local map and the global historical map, and updating the map feature information of the global historical map based on the difference information, includes:

[0025] Based on the target area, determine a local historical map in the global historical map that matches the current local map;

[0026] Based on the map element information of the local historical map and the map element information of the current local map, the difference information between the current local map and the local historical map is determined.

[0027] Based on the difference information and the local historical map, the map element information of the global historical map is updated.

[0028] In one embodiment, determining the difference information between the current local map and the local historical map based on the map feature information of the local historical map and the map feature information of the current local map includes:

[0029] Map element matching processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the one-to-one correspondence between the map elements contained in the current local map and the local historical map.

[0030] Map element difference processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the difference map element information between the current local map and the local historical map.

[0031] In one embodiment, the differential map feature information includes newly added map feature information, deleted map feature information, and changed map feature information. The map feature information also includes data collection time information of the structured data source corresponding to the map feature information.

[0032] Secondly, this application also provides a map updating device. The device includes:

[0033] The first acquisition module is used to acquire structured data sources within the target area;

[0034] The first determining module is used to determine the feature vector corresponding to the structured data acquisition data source based on preset data characteristics;

[0035] The second determining module is used to determine the accuracy probability of the structured data acquisition data source based on the feature vector corresponding to the structured data acquisition data source and the target big data model;

[0036] The generation module is used to generate a current local map describing real-world road information based on the target area and the structured data source, when the accuracy probability is higher than a preset accuracy probability threshold.

[0037] The update module determines the differences between the current local map and the global historical map, and updates the map element information of the global historical map based on the differences.

[0038] In one embodiment, the first determining module is specifically used for:

[0039] Based on the structured data source, determine the map feature data of the structured data source for the preset data features;

[0040] Based on the map feature data and the target coding strategy, determine the coded data corresponding to the map feature data;

[0041] Based on the encoded data, the feature vector corresponding to the structured acquisition data source is determined.

[0042] In one embodiment, the device further includes:

[0043] The second acquisition module is used to acquire a structured collection sample dataset, which includes multiple collection sample map data. The collection sample map data contains map feature information, data collection information, and feature true / false labels corresponding to the collection sample map data.

[0044] A construction module is used to determine the feature vector corresponding to each of the collected sample map data in the structured collected sample dataset, and to construct a target sample dataset based on the feature vector corresponding to each of the collected sample map data and the feature true / false labels;

[0045] The training module is used to input the target sample dataset into the initial big data model for model training to obtain the target big data model.

[0046] In one embodiment, the generation module is specifically used for:

[0047] If the accuracy probability is higher than a preset accuracy probability threshold, the structured data acquisition data source will be used as the input data source for the current local map data.

[0048] The input data source is subjected to multi-source data fusion processing to obtain the target map data;

[0049] Based on the target map data and the target area, a current local map describing real-world road information is generated.

[0050] In one embodiment, the update module is specifically used for:

[0051] Based on the target area, determine a local historical map in the global historical map that matches the current local map;

[0052] Based on the map element information of the local historical map and the map element information of the current local map, the difference information between the current local map and the local historical map is determined.

[0053] Based on the difference information and the local historical map, the map element information of the global historical map is updated.

[0054] In one embodiment, the update module is specifically used for:

[0055] Map element matching processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the one-to-one correspondence between the map elements contained in the current local map and the local historical map.

[0056] Map element difference processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the difference map element information between the current local map and the local historical map.

[0057] In one embodiment, the differential map feature information includes newly added map feature information, deleted map feature information, and changed map feature information. The map feature information also includes data collection time information of the structured data source corresponding to the map feature information.

[0058] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0059] Obtain structured data sources within the target area;

[0060] Based on preset data characteristics, determine the feature vector corresponding to the structured data acquisition data source;

[0061] Based on the feature vectors corresponding to the structured data source and the target big data model, determine the accuracy probability of the structured data source;

[0062] If the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source.

[0063] The differences between the current local map and the global historical map are determined, and the map feature information of the global historical map is updated based on the differences.

[0064] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0065] Obtain structured data sources within the target area;

[0066] Based on preset data characteristics, determine the feature vector corresponding to the structured data acquisition data source;

[0067] Based on the feature vectors corresponding to the structured data source and the target big data model, determine the accuracy probability of the structured data source;

[0068] If the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source.

[0069] The differences between the current local map and the global historical map are determined, and the map feature information of the global historical map is updated based on the differences.

[0070] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0071] Obtain structured data sources within the target area;

[0072] Based on preset data characteristics, determine the feature vector corresponding to the structured data acquisition data source;

[0073] Based on the feature vectors corresponding to the structured data source and the target big data model, determine the accuracy probability of the structured data source;

[0074] If the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source.

[0075] The differences between the current local map and the global historical map are determined, and the map feature information of the global historical map is updated based on the differences.

[0076] The aforementioned map update method, apparatus, computer equipment, storage medium, and computer program product acquire a structured data source within a target area; determine the feature vector corresponding to the structured data source based on preset data characteristics; determine the accuracy probability of the structured data source based on the feature vector and a target big data model; if the accuracy probability is higher than a preset accuracy probability threshold, generate a current local map describing real-world road information based on the target area and the structured data source; determine the difference information between the current local map and the global historical map; and update the map element information of the global historical map based on the difference information. This method can judge the accuracy probability of the structured data source based on the target big data model, and establish a current local map when the accuracy probability of the structured data source is higher than a preset accuracy probability threshold. Based on the current local map, the difference map element information to be updated can be determined, thereby improving the reliability of map element updates for autonomous driving maps. Attached Figure Description

[0077] Figure 1 This is an application environment diagram of the map update method in one embodiment;

[0078] Figure 2 This is a flowchart illustrating a map update method in one embodiment;

[0079] Figure 3 This is a schematic diagram of a semantic data representation model in one embodiment;

[0080] Figure 4 This is a schematic diagram illustrating the representation of each map element on a map in one embodiment;

[0081] Figure 5 This is a flowchart illustrating the process of determining the feature vector corresponding to a structured data acquisition data source in one embodiment.

[0082] Figure 6 This is a flowchart illustrating an example of constructing a feature vector corresponding to a structured data acquisition data source in one embodiment;

[0083] Figure 7 This is a flowchart illustrating the process of obtaining the target big data model in one embodiment;

[0084] Figure 8 This is a schematic diagram of the structure of an initial big data model in one embodiment;

[0085] Figure 9 This is a schematic diagram of the process for generating a current local map in one embodiment;

[0086] Figure 10This is a schematic diagram of a process for determining the difference information between a current local map and a local historical map in one embodiment;

[0087] Figure 11 This is a flowchart illustrating the process of determining the difference map element information between the current local map and the local historical map in one embodiment.

[0088] Figure 12 This is a structural block diagram of a map updating device in one embodiment;

[0089] Figure 13 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0090] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0091] The map update method provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices, such as smartwatches, smart bracelets, and head-mounted devices. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0092] In one embodiment, such as Figure 2 As shown, a map update method is provided, which can be applied to... Figure 1 Taking the server in the example, the following steps are included:

[0093] Step 202: Obtain the structured data source within the target area.

[0094] The structured data source is extracted from images captured by the terminal. Since images captured by devices typically contain a wealth of information—for example, a car's camera image might contain lane markings, signage, and lane-level speed limits—the car's terminal processes this image to obtain the corresponding structured data source. Therefore, each structured data source contains at least one piece of structured data. Each piece of structured data includes map feature information and data acquisition information. The map feature information consists of map attribute information and geometric information corresponding to the structured data. Geometric information includes the geometric representation of the structured data as points, lines, or polygons.

[0095] For example, map feature information may include geometric information of traffic signs, traffic sign symbols, square symbols, and the color blue. Data collection information may include, but is not limited to, the time of collection (i.e., solar altitude angle), date, vehicle model, collection environment information, the inertial navigation system installed in the vehicle, the model of the inertial navigation system installed in the vehicle, the model of the roadside equipment, whether the roadside equipment is calibrated, the calibration accuracy of the roadside equipment, the time synchronization accuracy, the number of semantic backhauls, and the collected real-time information, etc.

[0096] In this embodiment of the application, the server obtains the structured data source within the target area by receiving structured data sources collected by each terminal.

[0097] Each terminal can include an in-vehicle terminal installed in the vehicle and a terminal corresponding to the roadside equipment. The structured data acquisition source can be real-world images captured by the in-vehicle camera (e.g., a dashcam or in-vehicle camera) during vehicle operation or parking, and road environment data collected by vehicle sensors (e.g., LiDAR, millimeter-wave radar). The in-vehicle terminal then performs semantic recognition on the real-world images and road environment data. Alternatively, the structured data acquisition source can be real-world images captured by the roadside equipment's camera, and road environment data collected by the roadside equipment's sensors. The terminal corresponding to the roadside equipment then performs semantic recognition on the real-world images and road environment data. This application does not specifically limit the methods for obtaining semantic recognition and structured data acquisition sources.

[0098] Optionally, the server can construct a map data storage platform. Since the acquisition devices and accuracy of structured data sources differ, these data sources may be multi-source heterogeneous. This map data storage platform can store multi-source heterogeneous data provided by various users. It can also preprocess this multi-source heterogeneous data to achieve multi-source fusion and standardization. Users can include map administrators and individual users, and the multi-source heterogeneous data can include data from vehicle terminals, roadside equipment, and the cloud. This map data storage platform stores sufficient data sources, facilitating the server's updating of map elements in the global historical map based on the multi-source heterogeneous data from this platform. The map data storage platform can also autonomously update map elements in the global historical map based on the server it resides on.

[0099] It is understandable that the terminal corresponding to the vehicle or roadside equipment, after collecting the semantic data corresponding to the real-world images and road environment data, can store the semantic data as structured data, and then express this structured data according to the category of semantic data and the expression model, namely road-level information, lane-level information, and object information, as well as information describing the logical relationships between these information. The expression model can be referenced from [reference needed]. Figure 3 As shown.

[0100] The road-level information storage includes road geometry, road attributes, and road connectivity. Road geometry is the abstract vectorization of the road surface into a polyline (road reference line) with three-dimensional coordinates. Road-level attributes include, but are not limited to, road gradient, curvature, heading, traffic direction, number of lanes, road type, separation of up and down traffic, full enclosure, interchange, ramps, tunnels, bridges, overhead obstacles, rest areas, toll stations, toll plazas, checkpoints, intersections, roundabouts, U-turns, advance left turns, advance right turns, tidal flow roads, road-level traffic restrictions, road-level speed limits, road-level long-term construction, administrative divisions, and driving rules information. At the same time, the road connectivity is stored.

[0101] Lane-level information stores lane geometry, lane attributes, and lane connectivity, using polylines to represent lane centerlines. Lane-level attributes include, but are not limited to, lane number, direction of travel, lane type, lane width, intersection, lane-level traffic restrictions, lane-level speed limits, lane gradient, curvature, heading, lane markings (including but not limited to color, width, type, material, and traversability attributes), lane-level long-term construction, card issuance type, and payment method information. At the same time, it stores lane connectivity.

[0102] Object information stores the object's geometry and attribute information. Geometric representations include points, lines, and surfaces. Object attributes include, but are not limited to, object categories (including but not limited to curbs, guardrails, walls, poles, delineators, traffic lights, and pedestrian walkways), road surface markings (including but not limited to traffic signs and overhead obstacles), and road surface markings (including but not limited to arrows, text, symbols, warning zones, traffic dividers, stop locations, pedestrian crossings, public transport stops, and speed bumps).

[0103] Logical relationships can store the relationships between roads and lanes, and the association information between objects and lanes or roads.

[0104] These elements can be categorized into explicit elements and implicit elements. Explicit elements refer to real-world physical objects expressed in an abstract and digital form, while implicit elements refer to elements representing roads and their ancillary facilities through computational processing using explicit element information. Therefore, the explicit element updates involved in this application can be used for map application updates by various map managers, and the implicit element updates are performed by calculating the expression of explicit elements and update information to update the entire autonomous driving map.

[0105] Among them, the expression applicable to explicit elements is mainly object elements, including but not limited to: curbs, guardrails, poles, reflective markers, traffic lights, traffic signs, speed bumps, arrows, text, symbols, warning zones, traffic guidance zones, stop positions, pedestrian crossings, no-parking zones, etc.

[0106] When updating map features, refer to Figure 4 As shown, when each map feature information is expressed on the map, the map feature information may include, but is not limited to, feature geometry, feature attributes and update type; time domain; location accuracy confidence; existence confidence and classification correctness confidence.

[0107] Among them, the location accuracy confidence score is used to characterize the reliability information of the location accuracy collected by the structured data source; the existence confidence score is used to characterize the reliability information of whether the structured data source exists in the real world; and the classification correctness confidence score is used to characterize the reliability information of whether the map feature information of the structured data source is classified correctly.

[0108] Step 204: Determine the feature vector corresponding to the structured data acquisition data source based on the preset data characteristics.

[0109] In this embodiment of the application, the server may pre-store multiple preset data features. For example, the preset data features may include, but are not limited to, the time of collection (i.e., solar altitude angle), date, vehicle model, collection environment information, the inertial navigation system installed in the vehicle, the model of the inertial navigation system installed in the vehicle, the model of the roadside equipment, whether the roadside equipment is calibrated, the calibration accuracy of the roadside equipment, the time synchronization accuracy, the number of semantic backhauls, and the collected real-time information, etc.

[0110] Regarding the time of data collection, since the solar altitude angle affects the terminal's sensor's perception of the world, and time can directly and intuitively reflect the change in the solar altitude angle, this application can divide the influence of the solar altitude angle on sensor perception into a reasonable dimension by using a preset number of two-hour periods as a dimension. For example, for the time of data collection, the server uses two two-hour periods as an indicator to divide the time of a natural day (i.e., 24 hours) into 6-dimensional time information.

[0111] Regarding the collection date, since the real world appears differently depending on the season, the server divides the collection date into multi-dimensional date information.

[0112] Regarding vehicle models, different models are equipped with different types of sensors, and these sensors may have varying sensing capabilities. Therefore, the server differentiates between the different sensing capabilities of various sensors based on the vehicle model and categorizes the vehicle model into multi-dimensional vehicle information.

[0113] When collecting environmental information, changes in the environment can lead to different sensor readings, and extreme weather conditions can even cause sensors to malfunction. Therefore, the server divides different environmental information into multi-dimensional data, such as rain, snow, fog, and severe traffic congestion.

[0114] For inertial navigation systems (INS) on vehicles, the accuracy of data acquisition is higher when INS is installed, and high-precision map products have clear accuracy requirements. Therefore, the higher the accuracy of the collected data, the better the data quality. The server divides the vehicle's INS into multi-dimensional vehicle information.

[0115] Regarding the inertial navigation system (INS) model installed in a vehicle, since different models have varying levels of precision, higher precision in data acquisition results in better data quality. Therefore, the server can categorize vehicle information into multi-dimensional segments based on the INS model installed in the vehicle.

[0116] Different roadside equipment models are equipped with different sensors, and these sensors have varying sensing capabilities. Therefore, the server can differentiate the different sensing capabilities of different sensors based on the roadside equipment model and classify the roadside equipment model into multi-dimensional vehicle information.

[0117] Regarding whether roadside equipment is calibrated, calibrated roadside visual perception devices (i.e., cameras) collect data with higher accuracy, while data from uncalibrated roadside visual perception devices may be inaccurate. Therefore, the server uses the calibrated status of roadside equipment to define two-dimensional vehicle information.

[0118] The calibration accuracy and time synchronization accuracy of roadside equipment can be referred to the above description. The server will divide the calibration accuracy and time synchronization accuracy of roadside equipment into multi-dimensional vehicle information.

[0119] Regarding the number of semantic backtrackings, since the same structured data is collected multiple times, and the number of collections is a significant factor affecting the accuracy probability of the structured data source, the server can define semantic backtracking counts (i.e., the number of times it is collected) as 10 or more, and semantic backtracking counts as less than 10, into two-dimensional semantic backtracking counts.

[0120] For real-world scenarios, the collected map data varies greatly due to the differences in these scenarios, such as fully enclosed roads, urban roads, tunnels, slopes, curves, urban canyons, overhead obstacles, toll booths, and checkpoints. The server can divide the real-world scenario into multiple dimensions.

[0121] The dimensional division of each preset data feature can be set by technical personnel in actual application. This application does not make specific limitations on the number of preset data features and the granularity of division.

[0122] The preset data features can be obtained by the server through filtering various data collection information from the structured data source. Alternatively, the preset data features can be calculated by the server based on the indicator determination algorithm and various data collection information from the structured data source. This application does not limit the method for obtaining the preset data features.

[0123] The server obtains the feature vector corresponding to the structured data source based on the feature data of the structured data source for the preset data characteristics and the target encoding strategy.

[0124] Step 206: Determine the accuracy probability of the structured data source based on the feature vector corresponding to the structured data source and the target big data model.

[0125] In this embodiment of the application, the server inputs the feature vector corresponding to the structured data source into the target big data model, and the target big data model outputs the accuracy probability of the structured data source. Specifically, the target big data model can process each structured data in the structured data source in parallel, and thus simultaneously obtain the accuracy probability of each structured data in the structured data source.

[0126] The accuracy probability includes confidence in correct classification, confidence in location accuracy, and confidence in existence. The target big data model can be a big data model based on a fully connected neural network.

[0127] Step 208: If the accuracy probability is higher than the preset accuracy probability threshold, generate a current local map describing the real-world information of the road based on the target area and the structured data source.

[0128] In this embodiment, the server may pre-store preset accuracy probability thresholds. For example, when the accuracy probability includes classification correctness confidence, location accuracy confidence, and existence confidence, the server may pre-store preset accuracy probability thresholds corresponding to classification correctness confidence, location accuracy confidence, and existence confidence, respectively.

[0129] When the accuracy probability corresponding to the structured data source is higher than the preset accuracy probability threshold, the server creates a current local map within the target area, corresponding to the structured data source, to describe the real-world information of the road, based on the target area and the structured data source.

[0130] For example, the target area can be a 100m x 100m square area centered at longitude 116.4074 and latitude 39.9042. The map feature information corresponding to the structured data source includes traffic sign a, a dot, blue, longitude 116.4074, and latitude 39.9042. The server then creates an initial map based on the center point of the target area (longitude 116.4074, latitude 39.9042) and the map size (100m x 100m). The map feature information corresponding to the structured data source (i.e., traffic sign a, a dot, blue, longitude 116.4074, latitude 39.9042) is then expressed on the initial map to form a current local map.

[0131] Specifically, for each local map of the current situation, it should be constructed according to the scale of the corresponding global historical map. The shape of the local map of the current situation can also be circular; this application does not impose any restrictions on the shape of the local map of the current situation.

[0132] Step 210: Determine the differences between the current local map and the global historical map, and update the map feature information of the global historical map based on the differences.

[0133] In this embodiment, the server determines the differences between the map feature information in the current local map and the map feature information in the global historical map based on the map feature information in the current local map and the map feature information in the global historical map. Then, the server updates the map feature information in the global historical map based on the differences between the current local map and the global historical map.

[0134] In the aforementioned map update method, a structured data source within the target area is acquired; a feature vector corresponding to the structured data source is determined based on preset data characteristics; the accuracy probability of the structured data source is determined based on the feature vector and the target big data model; if the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source; the difference information between the current local map and the global historical map is determined; and the map element information of the global historical map is updated based on the difference information. This method can judge the accuracy probability of the structured data source based on the target big data model, and establish a current local map when the accuracy probability of the structured data source is higher than a preset accuracy probability threshold. Based on the current local map, the difference map element information to be updated can be determined, thus improving the reliability of map element updates for autonomous driving maps.

[0135] In one embodiment, such as Figure 5 As shown, step 204 includes:

[0136] Step 502: Based on the structured data source, determine the map feature data of the structured data source for the preset data features.

[0137] In this embodiment of the application, the server can determine the map feature data for each structured data source in the structured data source for each preset data feature based on a plurality of preset data features stored in advance.

[0138] Step 504: Determine the coded data corresponding to the map feature data based on the map feature data and the target coding strategy.

[0139] The target encoding strategy can be an encoding strategy based on ONEHOT (one-bit effective encoding). Any encoding strategy that can convert the categorical variables of discrete data into a form that is easy for machine learning algorithms to use can be used as the target encoding strategy. This application does not impose specific limitations on the target encoding strategy.

[0140] In the embodiments of the present application, the server may perform encoding processing on the map feature data corresponding to each piece of structured collected data in the structured collected data source based on the target encoding strategy, to obtain encoded data corresponding to each piece of map feature data.

[0141] For example, in a case where the map feature data includes a collection time period corresponding to the third and fourth of the twelve Earthly Branches, a device type being a roadside device, a calibration status of yes, and a semantic return count of 11 times, with reference to Figure 6 illustrated in the figure, the encoded data corresponding to the map feature data is obtained.

[0142] Step 506: Determine a feature vector corresponding to the structured collected data source according to the encoded data.

[0143] In the embodiments of the present application, for each piece of structured collected data in the structured collected data source, after obtaining the encoded data corresponding to the structured collected data, the server constructs the encoded data into a feature vector corresponding to the structured collected data. Then, the feature vector corresponding to the structured collected data source is obtained based on the feature vector corresponding to each piece of structured collected data in the structured collected data source.

[0144] For example, with reference to Figure 6 illustrated in the figure, the result of the server constructing the encoded data into the feature vector corresponding to the structured collected data is (1, 0, 1, 1, 1).

[0145] In this embodiment, since the map feature data corresponding to the structured collected data does not consider the mutual sequence and is independent of each other, the map feature data is relatively discrete. The server can perform normalization processing on the discrete map feature data through the target encoding strategy when the map feature data is relatively discrete, and the encoding strategy adopting ONEHOT encoding can also expand the number of features to obtain the feature vector. This facilitates subsequent input of the feature vector into the target big data model, so that the classifier of the target big data model can classify each feature vector more accurately, and obtain a more accurate accuracy probability result.

[0146] In one embodiment, as shown in Figure 7 illustrated in the figure, before step 206, the method further includes:

[0147] Step 702: Acquire a structured collected sample data set.

[0148] The structured collected sample data set includes a plurality of collected sample map data, and the collected sample map data includes map element information, data collection information, and a true / false label corresponding to the collected sample map data.

[0149] In this embodiment, the server can receive sample map data collected by terminals of a preset number of vehicles or roadside devices, and label each sample map data to determine the true or false labels of each sample map data. For each sample map data and its corresponding true or false label, the server uses that sample map data and its corresponding true or false label as collected sample map data. Then, the server constructs a structured collected sample dataset based on multiple collected sample map data.

[0150] It is understandable that, since the data sources of the map data of each sample in the structured sample dataset may be different, the map data of each sample in the structured sample dataset may be data with complete information on each map feature, or it may be data that is missing one or more map feature information.

[0151] The sample map data includes map feature information and data collection information. The true / false labels for each sample map data may include labels indicating correct classification, correct location, and correct existence, or may only include true / false labels for the entire map data set. This application does not specify a particular number of true / false labels.

[0152] Step 704: Determine the feature vectors corresponding to the map data of each collected sample in the structured sample dataset, and construct the target sample dataset based on the feature vectors and feature true / false labels corresponding to the map data of each collected sample.

[0153] In this embodiment, the server determines multiple map feature data for each sample map data in the structured acquisition sample dataset, based on preset data features. Then, the server determines multiple feature vectors corresponding to each map feature data according to the target encoding strategy.

[0154] The server can construct the target sample dataset from the feature vectors and true / false labels corresponding to each collected sample map data.

[0155] Step 706: Input the target sample dataset into the initial big data model for model training to obtain the target big data model.

[0156] In this embodiment of the application, the server inputs the target sample dataset into the initial big data model for model training and determines the accuracy of the model training. If the accuracy reaches the target accuracy threshold, the model is used as the target big data model.

[0157] The initial big data model can be constructed using a fully connected network, which consists of four layers, as shown in the reference. Figure 8As shown, the first layer is the input layer, the second and third layers are hidden layers, and the fourth layer is the output layer. The first layer (i.e., the input layer) is defined with 19 dimensions, the second hidden layer with 10 dimensions, the third hidden layer with 4 dimensions, and the output layer (i.e., the fourth layer) with 2 dimensions. The output obtained from the output layer can be mapped to the accurate probability using the softmax function (a mathematical function typically used to convert a set of arbitrary real numbers into real numbers representing a probability distribution). The target accuracy threshold can be determined by the technician in practical application; for example, the target accuracy threshold is 90%. This embodiment does not specifically limit the target accuracy threshold.

[0158] In this embodiment, because the map feature data corresponding to the initial sample map data are independent of each other and do not consider their order, the initial sample map feature data is relatively discrete. The server can normalize the discrete map feature data using a target encoding strategy, and the ONEHOT encoding strategy can further increase the number of features, resulting in feature vectors. Inputting the feature vectors of each initial sample map data into the initial big data model improves the iteration speed of the initial big data model, and the trained target big data model exhibits higher robustness.

[0159] In one embodiment, such as Figure 9 As shown, step 208 includes:

[0160] Step 902: If the accuracy probability is higher than the preset accuracy probability threshold, the structured data source is used as the input data source for the current local map data.

[0161] In this embodiment, if the accuracy probability of a certain structured data in the structured data source is higher than a preset accuracy probability threshold, the server uses that structured data as input data for the current local map data; if the accuracy probability of all structured data in the structured data source is higher than the preset accuracy probability threshold, the server uses the structured data source as input data for the current local map data.

[0162] The input data source contains multiple input data.

[0163] The server determines whether, before the current moment, there exists input data that matches the data collection information of the input data but has not undergone multi-source data fusion processing.

[0164] Step 904: Perform multi-source data fusion processing on the input data source to obtain the target map data.

[0165] The target map data includes map feature information and the accuracy probability corresponding to the target map data.

[0166] In this embodiment of the application, for any input data in the input data source, the server determines whether there is any input data that matches the data collection information of the input data before the current time and has not undergone multi-source data fusion processing. If there is no input data that matches the data collection information of the input data and has not undergone multi-source data fusion processing before the current time, the server treats the input data as input data that has not undergone multi-source data fusion processing.

[0167] If there is input data that matches the data collection information of the input data before the current time but has not undergone multi-source data fusion processing, the server will perform multi-source data fusion processing on the input data and the input data that matches the data collection information of the input data but has not undergone multi-source data fusion processing to obtain the target map data.

[0168] Specifically, the server uses a multi-source data fusion algorithm to perform multi-source data fusion processing on the input data from the input data source and the input data that matches the data collection information of the input data but has not undergone multi-source data fusion processing, in order to obtain the target map data.

[0169] For example, when the input data in the input data source and the map feature information of the input data that matches the data acquisition information of the input data but have not undergone multi-source data fusion processing differ only in positional accuracy by 5-10 cm, the server can perform multi-source data fusion processing on the input data in the input data source and the input data that matches the data acquisition information of the input data but have not undergone multi-source data fusion processing to obtain a target map data that balances the map feature information of the input data in the input data source and the input data that matches the data acquisition information of the input data but has not undergone multi-source data fusion processing.

[0170] Among them, for multi-source data fusion algorithms, any algorithm that can fuse map data that has certain differences in various dimensions to obtain target map data can be used in this application, and this application does not make any specific limitations on it.

[0171] Step 906: Generate a current local map describing real-world road information based on the target map data and the target area.

[0172] In this embodiment of the application, the server establishes a current local map within the target area corresponding to the target map data, based on the target area and the target map data.

[0173] The current local map includes map feature information corresponding to the target map data.

[0174] In this embodiment, when the accuracy probability of the structured data source is higher than a preset accuracy probability threshold, the server can perform multi-source fusion on the structured data source and the input data that matches the structured data source to obtain target map data with higher robustness and accuracy probability.

[0175] In one embodiment, such as Figure 10 As shown, step 210 includes:

[0176] Step 1002: Based on the target area, determine the local historical map that matches the current local map in the global historical map.

[0177] In this embodiment, the server can obtain a global historical map that requires map element updates. Based on the location information of the target area, a local historical map matching the current local map is determined within the global historical map. Specifically, if the target area is a 100m x 100m square centered at longitude 116.4074 and latitude 39.9042, the 100m x 100m square centered at longitude 116.4074 and latitude 39.9042 is determined within the global historical map and used as the local historical map.

[0178] Step 1004: Based on the map element information of the local historical map and the map element information of the current local map, determine the difference information between the current local map and the local historical map.

[0179] In this embodiment, the server determines the common information between the current local map and the local historical map based on the map element information of the local historical map and the map element information of the current local map. Then, the server determines the difference information between the current local map and the local historical map based on the common information between the current local map and the local historical map, the map element information of the local historical map, and the map element information of the current local map.

[0180] Step 1006: Based on the difference information and the local historical map, update the map feature information of the global historical map.

[0181] In this embodiment, the server uses the difference information between the current local map and the local historical map as the map element information to be updated in the local historical map, and then updates the map element information to be updated to the global historical map.

[0182] In this embodiment, the server can determine the difference information between the current local map and the local historical map that matches the current local map, and update the difference information to the global historical map, thereby achieving the effect of updating map elements on the static autonomous driving map.

[0183] In one embodiment, such as Figure 11 As shown, step 1004 includes:

[0184] Step 1102: Perform map element matching processing on the map element information of the local historical map and the map element information of the current local map to determine the one-to-one correspondence between the map elements contained in the current local map and the local historical map.

[0185] The one-to-one correspondence between the map elements contained in the current local map and the local historical map means that any map element in the current local map corresponds to any map element in the local historical map. The corresponding map element information indicates that the two map element information have only a slight difference in positional precision, while other attributes in the map element information must be completely identical. For example, a slight difference in positional precision can be 0.2 meters; this application does not limit the specific value of such a slight difference.

[0186] In this embodiment, the server invokes a preset map matching algorithm to match map feature information from the local historical map and the current local map, thereby obtaining matched map feature information from the local historical map and the current local map. The server can then use the matched map feature information from the local historical map and the current local map as corresponding map feature information.

[0187] Among them, any algorithm that can match map feature information between two maps can be applied to this application, and the embodiments of this application do not specifically limit it.

[0188] Step 1104: Perform map element difference processing on the map element information of the local historical map and the map element information of the current local map to determine the difference map element information between the current local map and the local historical map.

[0189] Then, the server invokes a preset map difference algorithm to perform map feature difference processing on the map feature information of the current local map and the map feature information of the local historical map, thereby obtaining map feature information that differs between the local historical map and the current local map. The server can use the map feature information that differs between the local historical map and the current local map as the difference map feature information.

[0190] Among them, any algorithm that can determine the differences between map feature information of two maps can be applied to the preset map difference algorithm, and the embodiments of this application do not specifically limit it.

[0191] In this embodiment, the server can match and differentiate the map feature information of the local historical map and the map feature information of the current local map to determine the difference map feature information between the local historical map and the current local map, so that the server can subsequently update the map features of the global historical map based on the difference map feature information.

[0192] In one embodiment, the differential map feature information includes newly added map feature information, deleted map feature information, and changed map feature information. The map feature information also includes data collection information from the structured data source corresponding to the map feature information.

[0193] In this embodiment of the application, the server can determine the update type of the differential map element information based on the specific differences between the current local map and the local historical map.

[0194] Specifically, the differences in map element information between the current local map and the local historical map are as follows: If a map element in the local historical map cannot be found in the current local map, the server determines that the update type of that map element is a deleted map element. If the color of a map element in the local historical map is different from the corresponding map element in the current local map, the server determines that the update type of that map element is a changed map element. If the update type of that map element is not found in the local historical map, the server determines that the update type of that map element is a newly added map element.

[0195] When the update type is "Add", the server can directly update the map feature information to the global historical map. When the update type is "Delete", the server determines the map feature information in the global historical map corresponding to the map feature information to be deleted, retrieves the map feature information that is the same as the map feature information in the global historical map, and deletes it. When the update type is "Change", the server retrieves the map feature information that is the same as the map feature information in the global historical map and updates the map feature information to the map. When the update type is "No Change", the server updates the confidence information of the map feature information in the global historical map, increasing the confidence of features that are confirmed to be unchanged and decreasing the confidence of map feature information whose status is uncertain.

[0196] The higher the confidence levels for location accuracy, existence, and correct classification, the higher the reliability of the representation; the lower the confidence levels, the lower the reliability.

[0197] It is understandable that for the same map feature information in the current local map and the local historical map, the server can determine that the change type of the same map feature information is the map feature information with no change.

[0198] For example, the map feature information for the difference between the current local map and the local historical map can be: feature geometry: point; feature attribute: traffic sign A, blue, solid line, 1px; update type: newly added map feature information; time domain: 10:08 to 14:07; location accuracy confidence: 90%; existence confidence: 95%; classification correctness confidence: 98%.

[0199] In this embodiment, when each map feature information is updated to the global historical map, the map feature information includes feature geometry, feature attributes, update type, newly added type map feature information, time domain, location accuracy confidence, existence confidence, and classification correctness confidence. Since the global historical map can display the time domain, location accuracy confidence, existence confidence, and classification correctness confidence corresponding to the map feature information after the map feature is updated, users can determine the freshness of the map feature information. Furthermore, the location accuracy confidence, existence confidence, and classification correctness confidence can also express the accuracy of the map feature information, which can provide feature and reliable confidence information for autonomous driving maps and enable large-scale and rapid updates of autonomous driving maps.

[0200] Understandably, since the map element information contained in the current local map is the map element information after multi-source fusion processing of structured data sources collected in a short period of time and at high frequency, the server can perform map element processing on different versions and specifications of autonomous driving maps (i.e. global historical maps) developed by any map administrator based on this current local map, thus achieving high universality of map element updates.

[0201] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0202] Based on the same inventive concept, this application also provides a map updating apparatus for implementing the map updating method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more map updating apparatus embodiments provided below can be found in the limitations of the map updating method described above, and will not be repeated here.

[0203] In one embodiment, such as Figure 12 As shown, a map updating device 1200 is provided, including: a first acquisition module 1202, a first determination module 1204, a second determination module 1206, a generation module 1208, and an update module 1210, wherein:

[0204] The first acquisition module 1202 is used to acquire structured data sources within the target area;

[0205] The first determining module 1204 is used to determine the feature vector corresponding to the structured data acquisition data source based on preset data characteristics;

[0206] The second determining module 1206 is used to determine the accuracy probability of the structured data acquisition data source based on the feature vector corresponding to the structured data acquisition data source and the target big data model.

[0207] The generation module 1208 is used to generate a current local map describing real-world road information based on the target area and the structured data source when the accuracy probability is higher than a preset accuracy probability threshold.

[0208] The update module 1210 determines the difference information between the current local map and the global historical map, and updates the map element information of the global historical map based on the difference information.

[0209] The map update device provided in this embodiment judges the accuracy probability of the structured data source based on the target big data model, and establishes a current local map when the accuracy probability of the structured data source is higher than a preset accuracy probability threshold. Based on the current local map, the differential map element information to be updated can be determined, which can improve the reliability of map element updates for autonomous driving maps.

[0210] In one embodiment, the first determining module 1204 is specifically used for:

[0211] Based on the structured data source, determine the map feature data of the structured data source for the preset data features;

[0212] Based on the map feature data and the target coding strategy, determine the coded data corresponding to the map feature data;

[0213] Based on the encoded data, the feature vector corresponding to the structured acquisition data source is determined.

[0214] In one embodiment, the apparatus further includes:

[0215] The second acquisition module is used to acquire a structured collection sample dataset, which includes multiple collection sample map data. The collection sample map data contains map feature information, data collection information, and feature true / false labels corresponding to the collection sample map data.

[0216] A construction module is used to determine the feature vector corresponding to each of the collected sample map data in the structured collected sample dataset, and to construct a target sample dataset based on the feature vector corresponding to each of the collected sample map data and the feature true / false labels;

[0217] The training module is used to input the target sample dataset into the initial big data model for model training to obtain the target big data model.

[0218] In one embodiment, the generation module 1208 is specifically used for:

[0219] If the accuracy probability is higher than a preset accuracy probability threshold, the structured data acquisition data source will be used as the input data source for the current local map data.

[0220] The input data source is subjected to multi-source data fusion processing to obtain the target map data;

[0221] Based on the target map data and the target area, a current local map describing real-world road information is generated.

[0222] In one embodiment, the update module 1210 is specifically used for:

[0223] Based on the target area, determine a local historical map in the global historical map that matches the current local map;

[0224] Based on the map element information of the local historical map and the map element information of the current local map, the difference information between the current local map and the local historical map is determined.

[0225] Based on the difference information and the local historical map, the map element information of the global historical map is updated.

[0226] In one embodiment, the update module 1210 is specifically used for:

[0227] Map element matching processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the one-to-one correspondence between the map elements contained in the current local map and the local historical map.

[0228] Map element difference processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the difference map element information between the current local map and the local historical map.

[0229] In one embodiment, the differential map feature information includes newly added map feature information, deleted map feature information, and changed map feature information. The map feature information also includes data collection time information of the structured data source corresponding to the map feature information.

[0230] Each module in the aforementioned map update device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0231] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 13As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data related to the various map update methods described above. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a map update method.

[0232] Those skilled in the art will understand that Figure 13 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0233] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0234] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0235] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0236] 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, data stored, data displayed, 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 the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0237] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0238] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0239] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A map updating method, characterized in that, The method includes: Obtain structured data sources within the target area; Based on preset data characteristics, determine the feature vector corresponding to the structured data acquisition data source; Based on the feature vectors corresponding to the structured data source and the target big data model, determine the accuracy probability of the structured data source; If the accuracy probability is higher than a preset accuracy probability threshold, a current local map describing real-world road information is generated based on the target area and the structured data source. The system identifies the differences between the current local map and the global historical map. Based on these differences, it updates the map element information of the global historical map. The map element information includes explicit elements and implicit elements. Explicit elements refer to real-world entities expressed in an abstract and digital form, while implicit elements refer to attribute elements of roads and their ancillary facilities expressed through computational processing using explicit element information. The explicit elements are used for map application updates by map administrators, and the updates of the implicit elements are performed through the expression and calculation of the explicit elements and update information, thereby updating the entire autonomous driving map. When the accuracy probability is higher than a preset accuracy probability threshold, the process of generating a current local map describing real-world road information based on the target area and the structured data source includes: If the accuracy probability is higher than a preset accuracy probability threshold, the structured data acquisition data source will be used as the input data source for the current local map data. Determine whether there exists historical input data before the current moment that matches the data acquisition information of any input data in the input data source and has not yet undergone multi-source data fusion processing; If the historical input data exists, the input data and the historical input data are subjected to multi-source data fusion processing to obtain the target map data; Based on the target map data and the target area, generate a current local map describing real-world road information; The method further includes: When the update type is "no change", the confidence information of the map element information in the global historical map is updated. The confidence of elements that are confirmed to have no change is increased, and the confidence of map element information whose change is uncertain is decreased. After the map element is updated, the global historical map displays the time domain, location accuracy confidence, existence confidence, and classification correctness confidence of the map element information.

2. The method according to claim 1, characterized in that, Determining the feature vector corresponding to the structured acquisition data source includes: Based on the structured data source, determine the map feature data of the structured data source for the preset data features; Based on the map feature data and the target coding strategy, determine the coded data corresponding to the map feature data; Based on the encoded data, the feature vector corresponding to the structured acquisition data source is determined.

3. The method according to claim 1, characterized in that, Before determining the accuracy probability of the structured data source based on the feature vector corresponding to the structured data source and the target big data model, the method further includes: Obtain a structured sample dataset, which includes multiple sample map data. The sample map data contains map feature information, data collection information, and feature true / false labels corresponding to the sample map data. Determine the feature vector corresponding to each of the collected sample map data in the structured collected sample dataset, and construct the target sample dataset based on the feature vector corresponding to each of the collected sample map data and the true / false labels of the features; The target sample dataset is input into the initial big data model for model training to obtain the target big data model.

4. The method according to claim 1, characterized in that, The process of determining the difference information between the current local map and the global historical map, and updating the map feature information of the global historical map based on the difference information, includes: Based on the target area, determine a local historical map in the global historical map that matches the current local map; Based on the map element information of the local historical map and the map element information of the current local map, the difference information between the current local map and the local historical map is determined. Based on the difference information and the local historical map, the map element information of the global historical map is updated.

5. The method according to claim 4, characterized in that, The determination of differences between the current local map and the local historical map based on map feature information from the local historical map and map feature information from the current local map includes: Map element matching processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the one-to-one correspondence between the map elements contained in the current local map and the local historical map. Map element difference processing is performed on the map element information of the local historical map and the map element information of the current local map to determine the difference map element information between the current local map and the local historical map.

6. The method according to claim 5, characterized in that, The differential map feature information includes newly added map feature information, deleted map feature information, and changed map feature information. The map feature information also includes data collection time information of the structured data source corresponding to the map feature information.

7. A map updating device, characterized in that, The device includes: The first acquisition module is used to acquire structured data sources within the target area; The first determining module is used to determine the feature vector corresponding to the structured data acquisition data source based on preset data characteristics; The second determining module is used to determine the accuracy probability of the structured data acquisition data source based on the feature vector corresponding to the structured data acquisition data source and the target big data model; The generation module is used to generate a current local map describing real-world road information based on the target area and the structured data source, when the accuracy probability is higher than a preset accuracy probability threshold. The update module is used to determine the differences between the current local map and the global historical map, and to update the map element information of the global historical map based on the differences. The map element information includes explicit elements and implicit elements. Explicit elements refer to real-world physical objects expressed in an abstract and digital form, while implicit elements refer to attribute elements of roads and their ancillary facilities expressed through calculation using explicit element information. The explicit elements are used for map application updates by map administrators, and the updates of the implicit elements are performed through the expression and calculation of the explicit elements and update information, thereby updating the entire autonomous driving map. The generation module is specifically used to use the structured data acquisition data source as the input data source for the current local map data when the accuracy probability is higher than the preset accuracy probability threshold. Determine whether there exists historical input data before the current moment that matches the data acquisition information of any input data in the input data source and has not yet undergone multi-source data fusion processing; If the historical input data exists, the input data and the historical input data are subjected to multi-source data fusion processing to obtain the target map data; Based on the target map data and the target area, generate a current local map describing real-world road information; The update module is also used to update the confidence information of the map element information in the global historical map when the update type is no change type, increase the confidence of the elements that are confirmed to have no change, and decrease the confidence of the map element information whose changes are uncertain; after the map elements are updated, the global historical map displays the time domain, location accuracy confidence, existence confidence and classification correctness confidence of the map element information.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images