Road semantic element generation method, system and device and medium

Abstract

The embodiment of the invention provides a road semantic element generation method. The road semantic element generation method comprises the following steps: acquiring original road data of a target area; extracting an initial road semantic element set of the target area from the original road data, wherein the initial road semantic element set comprises a plurality of initial road semantic elements; obtaining a reference road element set related to the target area, wherein the reference road element set comprises a plurality of reference road elements; the initial road semantic element set is optimized according to the reference road element set, a target road semantic element set of the target area is obtained, the target road semantic element set comprises multiple target road semantic elements, and the precision of the multiple target road semantic elements is larger than that of the multiple reference road elements.

Description

Technical Field

[0001] The present specification relates to the field of mobility control, and in particular to a method, system, device and medium for generating road semantic elements. Background Art

[0002] For mobile control of terminals, for example, when vehicles are performing intelligent driving, they usually need to rely on road semantic elements for auxiliary perception, planning, and decision-making. Usually, algorithms are used to automatically extract road semantic elements. However, the actual road conditions are complex and changeable. There may be situations such as ground markings being stained or redrawn, which will cause traditional algorithms to be unable to reliably detect road semantic elements in the environment.

[0003] Therefore, it is hoped to propose a road semantic element generation method, system, device and medium that can automatically extract road semantic elements more accurately and efficiently to adapt to complex road conditions. Summary of the invention

[0004] One of the embodiments of the present specification provides a method for generating road semantic elements, the method comprising: obtaining original road data of a target area; extracting an initial road semantic element set of the target area from the original road data, the initial road semantic element set comprising a plurality of initial road semantic elements; obtaining a reference road element set related to the target area, the reference road element set comprising a plurality of reference road elements; optimizing the initial road semantic element set according to the reference road element set to obtain a target road semantic element set of the target area, the target road semantic element set comprising a plurality of target road semantic elements, the accuracy of the plurality of target road semantic elements being greater than the accuracy of the plurality of reference road elements.

[0005] One of the embodiments of the present specification provides a road semantic element generation system, which includes: a first acquisition module, used to acquire original road data of a target area; an extraction module, used to extract an initial road semantic element set of the target area from the original road data, wherein the initial road semantic element set includes multiple initial road semantic elements; a second acquisition module, used to acquire a reference road element set related to the target area, wherein the reference road element set includes multiple reference road elements; an optimization module, used to optimize the initial road semantic element set according to the reference road element set to obtain a target road semantic element set of the target area, wherein the target road semantic element set includes multiple target road semantic elements, and the accuracy of the multiple target road semantic elements is greater than the accuracy of the multiple reference road elements.

[0006] One of the embodiments of the present specification provides a road semantic element generation device, including a processor, wherein the processor is used to execute the road semantic element generation method as described in the above embodiment.

[0007] One of the embodiments of the present specification provides a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the road semantic element generation method as described in the above embodiment. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0009] Figure 1 is a schematic diagram of an application scenario of a road semantic element generation system according to some embodiments of this specification;

[0010] Figure 2 is an exemplary module diagram of a road semantic element generation system according to some embodiments of this specification;

[0011] Figure 3 is an exemplary flow chart of a method for generating road semantic elements according to some embodiments of this specification;

[0012] Figure 4 is an exemplary flow chart of determining an initial road semantic element set according to some embodiments of this specification;

[0013] Figure 5 is an exemplary flow chart of determining a target road semantic element set of a target area according to some embodiments of this specification;

[0014] Figure 6 is another exemplary flow chart of determining a target road semantic element according to some embodiments of this specification;

[0015] Fig. 7A is a schematic diagram of an image frame acquired by capturing a target area by a reference vehicle according to some embodiments of this specification;

[0016] Figure 7B is a schematic diagram of another image frame acquired by capturing a target area using a reference vehicle according to some embodiments of this specification;

[0017] Figure 7C is a schematic diagram of another image frame acquired by capturing a target area using a reference vehicle according to some embodiments of this specification;

[0018] Fig. 8A is a schematic diagram of a reference map according to some embodiments of this specification;

[0019] Figure 8B is a schematic diagram of a first topology block according to some embodiments of the present specification. DETAILED DESCRIPTION

[0020] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0021] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0022] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0023] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0024] Figure 1 It is a schematic diagram of an application scenario of a road semantic element generation system according to some embodiments of this specification.

[0025] The systems and methods in this specification can be applied to scenarios of terminal mobile control, for example, autonomous driving scenarios (e.g., pure unmanned driving without a driver); for example, scenarios such as assisted driving in vehicles. The assisted driving module in the vehicle can autonomously control the vehicle's driving, and can also output driving prompts to the driver. Movable terminals can be applied to land, sea, aerospace, etc., or any combination thereof. Movable terminals may include taxis, private cars, ride-sharing, shared vehicles, buses, trains, motor vehicles, high-speed railways, subways, ships, airplanes, spacecraft, hot air balloons, bicycles, tricycles, motorcycles, robots, etc., or any combination thereof. In some embodiments, the road semantic element generation system can generate road semantic elements of the target area by implementing the methods and / or processes disclosed in this specification to achieve mobile control of the terminal, for example, performing operations such as perception, positioning, prediction, decision-making and planning related to the movement of the terminal.

[0026] like Figure 1 As shown, the application scenario 100 of the road semantic element generation system may include a target area 110 , original road data 120 , a reference map 130 , a network 140 , a server 150 , and a terminal 160 .

[0027] The target area 110 is a road section where road semantic elements need to be generated.

[0028] The original road data 120 is data related to the road in the target area 110. The original road data 120 may include, but is not limited to, one or more of point cloud data (e.g., point cloud frames), image data (e.g., image frames), time information, location information, speed, acceleration, and direction related to the road in the target area 110. Figure 1 As shown, the original road data 120 can be obtained by a collection device (such as a radar, a camera, a speed sensor, a positioning component, etc.) on a reference vehicle 121. The aforementioned reference vehicle 121 can refer to a vehicle that collects road-related data. For more information about the target area and the original road data, please refer to Figure 3 and its related description.

[0029] The reference map 130 is data related to the roads currently available in the target area 110. For example, the reference map 130 may be a precision map corresponding to the target area 110.

[0030] The network 140 can connect the components of the application scenario 100 of the road semantic element generation system and / or connect the application scenario 100 of the road semantic element generation system with external resources. The application scenario 100 of the road semantic element generation system enables communication between the components and other components outside the application scenario 100 of the road semantic element generation system to promote the exchange of data and / or information. For example, the reference vehicle 121 can transmit the original road data 120 to the server 150 through the network 140. For another example, the reference vehicle 121 can also obtain the relevant instructions issued by the server 150 through the network 140. For another example, the terminal 160 can obtain the target road semantic element set of the target area issued by the server 150 through the network 140. For more information about the above examples, please refer to the following of this specification. The network 140 can be any one or more of a wired network or a wireless network. For example, the network 140 can include but is not limited to a cable network, an optical fiber network, a telecommunications network, the Internet, a Bluetooth network, a near field communication, a bus within a device, a line within a device, a cable connection, etc. or any combination thereof. The network connection between the various parts can be in one of the above methods or in multiple methods. The network 140 may be of any topology, such as point-to-point, shared, centralized, or a combination of multiple topologies.

[0031] The server 150 is a system with computing capabilities. The server 150 can process data and / or information obtained from various components of the application scenario 100 of the road semantic element generation system or other devices, and execute program instructions based on these data, information and / or processing results to perform one or more functions described in this specification. For example, the server 150 can obtain original road data of the target area; extract an initial road semantic element set of the target area from the original road data, and the initial road semantic element set includes multiple initial road semantic elements; obtain a reference road element set related to the target area, and the reference road element set includes multiple reference road elements; optimize the initial road semantic element set according to the reference road element set to obtain a target road semantic element set of the target area, and the target road semantic element set includes multiple target road semantic elements, and the accuracy of the multiple target road semantic elements is greater than the accuracy of the multiple reference road elements. For more information about the above embodiments, please refer to Figure 3and its related description. The server 150 may include one or more processing devices (e.g., a single-core processing device or a multi-core multi-core processing device). As an example only, the server 150 may include a central processing unit, a dedicated integrated circuit, a dedicated instruction processor, a graphics processor, a physical processor, a microprocessor, etc., or any combination thereof. In some embodiments, the server 150 may be integrated into at least one component in the application scenario 100 of the road semantic element generation system, for example, the terminal 160.

[0032] The terminal 160 may refer to a terminal device used by a user associated with the target area 110. The aforementioned user may refer to a person who may pass through the target area 110. Figure 1 As shown, the aforementioned terminal device may include but is not limited to a mobile phone, a tablet, a computer or a vehicle. The aforementioned vehicle may be the reference vehicle 121, or may be another intelligent driving vehicle. The aforementioned intelligent driving vehicle may include but is not limited to autonomous driving assistance vehicles, networked driving assistance vehicles, human-machine co-driving vehicles and autonomous driving vehicles. The aforementioned terminal device may also include other devices, for example, a vehicle computer on a vehicle.

[0033] As shown in some embodiments of this specification, in an application scenario 100 of a road semantic element generation system, a server 150 can obtain original road data 120 of a target area 110 through a network 140, and the server 150 can obtain a reference map 130 of the target area 110 through the network 140, and generate a target road semantic element set of the target area 110 based on the original road data 120 and the reference map 130, thereby realizing automatic extraction of road semantic elements of the target area 110 to adapt to complex road conditions. For more information about the above embodiments, please refer to Figure 3 and its related description.

[0034] It should be noted that the application scenario 100 in the road semantic element generation system is provided for illustrative purposes only and is not intended to limit the scope of this specification. For those of ordinary skill in the art, various modifications or changes can be made based on the description of this specification. For example, the application scenario 100 in the road semantic element generation system may also include a storage device, a database, an information source, and the like. For another example, the application scenario 100 in the road semantic element generation system may be implemented on other devices to achieve similar or different functions. However, changes and modifications will not deviate from the scope of this specification.

[0035] Figure 2 2 is an exemplary module diagram of a road semantic element generation system according to some embodiments of the present specification. The road semantic element generation system 200 may be implemented by the server 150 and / or the terminal 160 .

[0036] like Figure 2As shown, the road semantic element generation system 200 may include a first acquisition module 210 , an extraction module 220 , a second acquisition module 230 and an optimization module 240 .

[0037] The first acquisition module 210 can be used to acquire the original road data of the target area. In some embodiments, the original road data of the target area is divided into multiple original data sets, each of which includes the original road data obtained by detecting the target area at the detection time. For each original data set, the first acquisition module 210 can be further used to process the original road data in the original data set based on a semantic detection model to determine the initial road sub-semantic elements corresponding to the original data set, wherein the semantic detection model is a machine learning model; based on the geometric relationship and attribute characteristics of the initial road sub-semantic elements corresponding to the multiple original data sets, the initial road sub-semantic elements corresponding to the multiple original data sets are divided into multiple candidate road semantic element groups; the initial road sub-semantic elements in each candidate road semantic element group are merged to determine the multiple initial road semantic elements; based on the multiple initial road semantic elements, the initial road semantic element set is determined.

[0038] The extraction module 220 may be configured to extract an initial road semantic element set of the target area from the original road data, wherein the initial road semantic element set includes a plurality of initial road semantic elements.

[0039] The second acquisition module 230 may be configured to acquire a reference road element set related to the target area, wherein the reference road element set includes a plurality of reference road elements.

[0040] The optimization module 240 can be used to optimize the initial road semantic element set according to the reference road element set to obtain the target road semantic element set of the target area, wherein the target road semantic element set includes multiple target road semantic elements, and the accuracy of the multiple target road semantic elements is greater than the accuracy of the multiple reference road elements. In some embodiments, the optimization module 240 can be further used to determine a first road structure change point based on the reference road element set; and determine multiple first topological blocks according to the first road structure change point. Determine a second road structure change point according to the initial road semantic element set; determine multiple second topological blocks according to the second road structure change point; for each second topological block, determine the first topological block that matches the second topological block; optimize the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block that matches the second topological block to obtain the target road semantic element of the second topological block; and determine the target road semantic element set of the target area based on the target road semantic elements corresponding to the multiple second topological blocks. In some embodiments, the optimization module 240 may be further used to process the initial road semantic element corresponding to the second topological block and the reference road element corresponding to the first topological block matching the second topological block based on the optimization model to determine the target road semantic element corresponding to the second topological block. In some embodiments, the optimization module 240 may be further used to determine the road semantic element to be optimized corresponding to the second topological block based on the initial road semantic element corresponding to the second topological block; optimize the road semantic element to be optimized based on the reference road element corresponding to the first topological block matching the second topological block to determine the target road semantic element corresponding to the second topological block. In some embodiments, the optimization module 240 may be further used to determine the road semantic element to be optimized corresponding to the second topological block based on the initial road semantic element corresponding to the second topological block; determine whether the road semantic element to be optimized needs to meet the preset accuracy requirement; in response to determining that the road semantic element to be optimized does not need to meet the preset accuracy requirement, determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block; and determine the reference road element corresponding to the road semantic element to be optimized as the target road semantic element.

[0041] For more information about the above modules, please refer to the relevant content below in this manual.

[0042] It should be noted that the above description of the road semantic element generation system 200 and its modules is only for convenience of description and does not limit the present specification to the scope of the embodiments. It is understandable that, after understanding the principle of the system, those skilled in the art may arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from the principle. In some embodiments, Figure 1 The first acquisition module 210, extraction module 220, second acquisition module 230 and optimization module 240 disclosed in the specification can be different modules in a system, or one module can realize the functions of two or more modules. For example, each module can share a storage module, or each module can have its own storage module. Such variations are within the protection scope of this specification.

[0043] The following description of the generation of road semantic elements takes autonomous driving as an example, but it can be understood that the method and system for generating road semantic elements are also applicable to other mobile control scenarios, such as robots.

[0044] Figure 3 is an exemplary flow chart of a method for generating road semantic elements according to some embodiments of the present specification. In some embodiments, process 300 may be executed by server 150 and / or terminal 160. For example, process 300 may be executed by road semantic element generation system 200 implemented by server 150 and / or terminal 160.

[0045] The process 300 may be used to generate road semantic elements online or offline.

[0046] In some embodiments, in an autonomous driving scenario, the road semantic element generation method described in process 300 can be generated online, and the generated target road semantic elements can be subsequently used for mobile control of the intelligent driving vehicle, and can also be used for the automated production of high-precision road semantic elements. The aforementioned process 300 of generating road semantic elements online can be executed by the server 150 or the intelligent driving vehicle.

[0047] For example, the smart driving vehicle can detect the original road data and send it to the server 150. The server 150 executes 300 to generate the target road semantic element and then sends it back to the smart driving vehicle, thereby ensuring that the smart driving vehicle can perform movement control according to the target road semantic element.

[0048] For another example, the intelligent driving vehicle can detect the original road data, and the processing component of the intelligent driving vehicle itself executes 300 to generate the target road semantic element, so as to ensure that the intelligent driving vehicle can perform movement control according to the target road semantic element. The generated target road semantic element can also be used for the generation or update of high-precision road elements (for example, high-precision maps).

[0049] In some embodiments, the method for generating road semantic elements described in process 300 may be generated offline, and the generated target road semantic elements may be subsequently used to generate or update high-precision road elements (e.g., high-precision maps). The process 300 of generating road semantic elements offline may be executed by the server 150.

[0050] For example, the intelligent driving vehicle can detect the original road data and send it to the server 150. The server 150 executes process 300 to generate target road semantic elements for the generation or update of high-precision road elements (e.g., high-precision maps).

[0051] like Figure 3 As shown, process 300 includes the following steps:

[0052] Step 310: Acquire original road data of the target area.

[0053] The first acquisition module 210 may acquire original road data of the target area collected by the reference vehicle 121 .

[0054] For example, in the scenario of online generation of road semantic elements, the target area may be a certain range around the current position of the reference vehicle 121. The reference vehicle 121 may continuously collect data on the target area based on a collection device disposed thereon, thereby obtaining original road data of the target area.

[0055] For another example, in the scenario of generating road semantic elements offline, the server 150 may pre-collect the road data uploaded by the reference vehicle 121, and select the original road data related to the preset area (ie, the target area) from the aforementioned road data for processing.

[0056] Step 320: extracting an initial road semantic element set of the target area from the original road data.

[0057] The initial road semantic element set is a collection of initial road semantic elements. The initial road semantic elements may refer to the road semantic elements in the target area determined based on the original road data. The aforementioned road semantic elements may refer to the road elements and their attribute characteristics related to the road determined based on the original road data. The aforementioned road elements may include pavement elements related to the road surface and roadside elements related to the roadside, wherein the aforementioned pavement elements may include but are not limited to lane lines, stop lines, lane markings, road boundaries, sidewalks, etc., and the aforementioned roadside elements may include traffic lights, street lights, signs, etc. The attribute characteristics of the aforementioned road elements may include but are not limited to one or more of location, category, shape, etc. For example, the road semantic elements may include the type and location of lane markings.

[0058] The extraction module 220 may perform modeling or adopt various data analysis algorithms, such as regression analysis, discriminant analysis, etc., to analyze and process the original road data and extract an initial road semantic element set of the target area.

[0059] In some embodiments, for each of the original data sets, the extraction module 220 can also process the original road data in the original data set based on a semantic detection model to determine the initial road sub-semantic elements corresponding to the original data set, wherein the semantic detection model is a machine learning model; based on the geometric relationship and attribute characteristics of the initial road sub-semantic elements corresponding to the multiple original data sets, the initial road sub-semantic elements corresponding to the multiple original data sets are divided into multiple candidate road semantic element groups; the initial road sub-semantic elements in each of the candidate road semantic element groups are merged to determine the multiple initial road semantic elements; based on the multiple initial road semantic elements, the initial road semantic element set is determined. For more information about the aforementioned embodiments, please refer to Figure 4 and its related description.

[0060] Step 330: Acquire a reference road element set related to the target area.

[0061] The reference road element set is a collection of reference road elements. The reference road element may refer to an element related to road information in a target area determined based on a reference map. Similar to the road semantic element, the reference road element may also include attribute features of each road element, and the attribute features of each road element in the reference road element may also include but are not limited to position, category, shape, etc. It can be understood that since the reference map is an existing map for reference, its accuracy may be lower than the accuracy of each road element in the road semantic element. For example, the accuracy of the position of the lane line in the aforementioned road semantic element is centimeter-level accuracy, while the accuracy in the reference road element is meter-level accuracy. In addition, the reference road element may also include attribute information of the road. For example, the reference road element may also include speed limits, bus lane information, traffic restriction information, etc. of the lanes in the road.

[0062] The second acquisition module 230 can obtain the location information of the reference vehicle 121, and based on the aforementioned location information, obtain a reference map within a preset range related to the target area. For example, the second acquisition module 230 can obtain a reference map within the target area and a range of 10 meters extending outward. The aforementioned preset range can be related to the accuracy of the reference map. The higher the accuracy of the reference map, the smaller the range corresponding to the reference map can be. It can be understood that since the reference map may have a certain error, if only the reference map corresponding to the target area is obtained, the aforementioned reference map obtained may not completely cover the actual target area due to the error. Therefore, some embodiments of this specification obtain a reference map within a preset range related to the target area to ensure that the reference map obtained covers the target area completely. The location information of the aforementioned reference vehicle 121 can be obtained in two modes: online and offline. In the online mode, the second acquisition module 230 can obtain the real-time scene through the on-site real-time positioning of the reference vehicle 121, and then determine the location information of the reference vehicle 121. In the offline mode, the position of the reference vehicle 121 can be optimized, for example, according to the global trajectory information of the reference vehicle 121 (e.g., based on Slam optimization). It is understandable that when locating a certain position of the reference vehicle 121, due to poor global positioning system (GPS) signals or other reasons, the initial position information may have certain errors, but its approximate position can be determined. At this time, if the reference vehicle 121 returns to the vicinity of the same reference vehicle 121 after circling a circle, the position information obtained at this time and the initial position information have certain errors, but since the two positions are actually the same position, this error can be used to optimize all the global data and all the positions on the global trajectory of the reference vehicle 121 to obtain the optimized position information, thereby improving the accuracy of the position information of the reference vehicle 121.

[0063] The second acquisition module 230 may analyze and process the reference map within a preset range related to the target area to determine a reference road element set related to the target area.

[0064] Step 340 , optimizing the initial road semantic element set according to the reference road element set to obtain a target road semantic element set of the target area.

[0065] The target road semantic element set may be a collection of target road semantic elements. The target road semantic element set may include multiple target road semantic elements. The target road semantic element refers to an optimized road semantic element, and the accuracy of multiple target road semantic elements is greater than the accuracy of multiple reference road elements. It is understandable that due to the complex and changeable road conditions in reality, there may be situations such as ground marking pollution and ground marking redrawing, which may cause the acquired initial road semantic element set to be missing or have low confidence. Some embodiments of this specification may optimize the initial road semantic element set according to the reference road element set to obtain the target road semantic element set of the target area.

[0066] In some embodiments, the optimization module 240 can perform modeling or adopt various data analysis algorithms, such as regression analysis, discriminant analysis, etc., to optimize the initial road semantic element set based on the reference road element set to obtain the target road semantic element set of the target area.

[0067] In some embodiments, the optimization module 240 can determine the first road structure change point based on the reference road element set; determine multiple first topological blocks based on the first road structure change point; determine the second road structure change point based on the initial road semantic element set; determine multiple second topological blocks based on the second road structure change point; for each second topological block, determine the first topological block that matches the second topological block; optimize the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block that matches the second topological block to obtain the target road semantic element of the second topological block; determine the target road semantic element set of the target area based on the target road semantic elements corresponding to the multiple second topological blocks. For more information about the aforementioned embodiments, please refer to Figure 5 and its related description.

[0068] Some embodiments of the present specification can determine an initial road semantic element set through original road data of a target area, and optimize the initial road semantic element set based on a reference road element set to obtain a target road semantic element set of the target area, thereby achieving better automated extraction of road semantic elements of the target area and adapting to complex road conditions.

[0069] Figure 4 is an exemplary flow chart of determining an initial road semantic element set according to some embodiments of this specification. In some embodiments, process 400 may be executed by extraction module 220. Figure 4 As shown, process 400 includes the following steps:

[0070] Step 410 : for each original data set, original road data in the original data set is processed based on the semantic detection model to determine the initial road sub-semantic element corresponding to the original data set.

[0071] In some embodiments, the original road data of the target area can be divided into multiple original data sets. The original data set can refer to a collection of original road data obtained at each detection time. Each original data set includes original road data obtained by detecting the target area at the detection time. Sub-areas corresponding to multiple original data sets can constitute the target area, and there may or may not be overlapping areas between the sub-areas corresponding to the multiple original data sets.

[0072] The detection time can be determined by preset, which can be a moment or a time period. For example, each raw data set may include point cloud data and image data related to the road at a moment or time period corresponding to the detection time. The laser point cloud in the reference vehicle 121 rotates a preset angle (for example, rotates 360°) to collect a frame of point cloud data; the camera in the reference vehicle 121 can rotate a preset angle (for example, rotates 360°) synchronously with the aforementioned laser radar to collect a frame of image data.

[0073] For example, when the reference vehicle continuously collects and obtains the original road data of the target area during each detection time, the detection time may correspond to the time period during which the reference vehicle continuously collects data. At this time, each original data set may include point cloud data and / or image data related to the road during the corresponding time period. For another example, when the reference vehicle collects and obtains the original road data of the target area at each preset detection time, the detection time may correspond to the moment when the reference vehicle collects data each time. At this time, each original data set may include point cloud data and / or image data related to the road at the corresponding moment (e.g., each original data set may include an image frame at the corresponding moment).

[0074] The initial road sub-semantic element may refer to a road semantic element in a sub-region corresponding to the original data set determined based on the original data set. For more information about the initial road sub-semantic element, see Figure 3 Related descriptions of road semantic elements.

[0075] The extraction module 220 may input the original data set into the semantic detection model, and the output of the semantic detection model may be the initial road sub-semantic elements in the sub-area corresponding to the original data set. The semantic detection model may be a machine learning model, for example, the semantic detection model may be constructed based on BEV fusion.

[0076] Exemplarily, the extraction module 220 may Figure 7B The image frame shown is input into the semantic detection model, and the output of the semantic detection model can be the initial road sub-semantic elements (such as lane width, position, number, type, and position and type of lane markings, etc.) within the sub-area corresponding to the image frame.

[0077] The semantic detection model can be acquired based on training, and the training sample can include a sample original data set, and the training label can include a sample initial road sub-semantic element obtained by manually annotating the aforementioned sample original data set. In some embodiments, the output of the semantic detection model also includes the confidence of the initial road sub-semantic element. Correspondingly, when training the semantic detection model, the training label can include the sample confidence (e.g., 1) manually assigned to the sample initial road sub-semantic element.

[0078] Step 420 : Based on the geometric relationships and attribute features of the initial road sub-semantic elements corresponding to the multiple original data sets, the initial road sub-semantic elements corresponding to the multiple original data sets are divided into multiple candidate road semantic element groups.

[0079] The candidate road semantic element group may include multiple initial road sub-semantic elements related to the same road semantic element at the same or similar position. Figure 7A to Figure 7C The lane markings and the attribute features corresponding to the lane markings are shown, wherein Fig. 7A The initial road sub-semantic element corresponding to the image frame shown includes the lane mark and its position B, category: straight sign, shape: width 0.6M*length 1.8M and other attribute features; Figure 7B as well as Figure 7C The initial road sub-semantic elements corresponding to the image frames shown include lane markings and the position B of the lane markings, categories: straight-ahead markings, shapes: width 0.6M*length 2M and other attribute features.

[0080] In some embodiments, the extraction module 220 may aggregate the initial road sub-semantic elements corresponding to the multiple original data sets to obtain at least one group of aggregated initial road sub-semantic elements. For example, when the extraction module 220 extracts the initial road sub-semantic elements for online detection, the extraction module 220 may use a sliding window to aggregate the initial road sub-semantic elements from the current detection time to a certain historical detection time to obtain at least one group of aggregated initial road sub-semantic elements. For another example, when the extraction module 220 extracts the initial road sub-semantic elements for offline, the extraction module 220 may segment the target area according to the road structure in the target area, and aggregate the initial road sub-semantic elements of each part of the target area to obtain at least one group of aggregated initial road sub-semantic elements.

[0081] In some embodiments, for each group of aggregated initial road sub-semantic elements, the extraction module 220 can compare the geometric relationship and attribute characteristics of each initial road sub-semantic element in the group of aggregated initial road sub-semantic elements, and divide each initial road sub-semantic element that meets the preset conditions into multiple candidate road semantic element groups. The geometric relationship may include but is not limited to shape similarity, position similarity, etc. The aforementioned geometric relationship can be determined based on the attribute characteristics of each initial road sub-semantic element. For example, the extraction module 220 can determine the shapes of lane markings corresponding to two initial road sub-semantic elements and compare them, thereby determining that the shape similarity of the two is 1 (i.e., exactly the same), and use it as the geometric relationship between the two initial road sub-semantic elements.

[0082] The aforementioned preset condition may be to determine whether each initial road sub-semantic element is related to the same road semantic element at the same or similar position. For example, the aforementioned preset condition may include that the geometric shape similarity and position similarity of multiple initial road sub-semantic elements are greater than a preset similarity threshold (e.g., 0.9), and the attribute features are completely identical. The extraction module 220 may divide two initial road sub-semantic elements whose shape similarity and position similarity are both 1 and whose attribute features are completely identical into a group of candidate road semantic elements.

[0083] Step 430: fuse the initial road sub-semantic elements in each candidate road semantic element group to determine a plurality of initial road semantic elements.

[0084] In some embodiments, for each candidate road semantic element group, the extraction module 220 may fuse the candidate road semantic element group to determine the fused initial road sub-semantic element. For each candidate road semantic element group, the extraction module 220 may determine the distribution type of the road element corresponding to the candidate road semantic element group based on a preset corresponding relationship, and determine the fusion method for fusing the candidate road semantic element group based on the aforementioned distribution type and through a preset fusion rule, and then fuse the candidate road semantic element group based on the aforementioned fusion method to determine the fused initial road sub-semantic element. The distribution type may be the type of distribution of the road element in the target area. The distribution type may include but is not limited to point distribution, line distribution, polygon distribution, etc. For example, traffic lights, lane signs, etc. are point distribution. For another example, lane lines are line distribution. The corresponding relationship may refer to the corresponding relationship between the road element and the distribution type. The fusion rule may refer to the corresponding rule between the distribution type and the fusion method. For example, the fusion method of the candidate road semantic element group corresponding to the point-distributed road elements includes Gaussian probability fusion and category voting algorithm. For another example, the fusion method for the candidate road semantic element group corresponding to the linearly distributed road elements is refitting. For another example, the fusion method for the candidate road semantic element group corresponding to the polygonally distributed road elements is union.

[0085] Exemplarily, the extraction module 220 can determine, based on a preset corresponding relationship, that the distribution type corresponding to a candidate road semantic element group related to traffic lights is a point distribution, and based on the point distribution, through a preset fusion rule, determine that the fusion method for fusing the candidate road semantic element group includes Gaussian probability fusion and a category voting algorithm. The extraction module 220 can calculate the position of the traffic light after fusion in the candidate road semantic element group based on Gaussian probability fusion, and determine other attribute features after the aforementioned traffic light fusion based on the category voting algorithm, thereby determining the fused initial road sub-semantic element.

[0086] For another example, the extraction module 220 can determine, based on a preset corresponding relationship, that the distribution type corresponding to a candidate road semantic element group related to the lane line is a linear distribution, and based on the linear distribution, through a preset fusion rule, determine that the fusion method for fusing the candidate road semantic element group is to first refit and then find the union. The extraction module 220 can determine the position of the traffic light after fusion in the candidate road semantic element group, and then determine other attribute features after the aforementioned traffic light fusion based on the category voting algorithm, thereby determining the initial road semantic element corresponding to the candidate road semantic element group.

[0087] In some embodiments, for each candidate road semantic element group, the extraction module 220 may directly determine the fused initial road sub-semantic element as the initial road semantic element corresponding to the candidate road semantic element group.

[0088] In some embodiments, for each candidate road semantic element group, the extraction module 220 can also perform rule filtering on the fused initial road sub-semantic elements to determine the initial road semantic elements corresponding to the candidate road semantic element group. The aforementioned rule filtering refers to setting a preset filtering rule to remove noise in the fused initial road sub-semantic elements and optimize the fusion result. For example, a candidate road semantic element group is related to a lane line. When a lane line with a length less than a preset threshold appears in the fused initial road sub-semantic element, or when the lane line overlaps with a crosswalk, or when the lane line exceeds the stop line, the extraction module 220 can remove the relevant initial road sub-semantic elements in the candidate road semantic element group by setting a preset filtering rule, and re-fuse until the fused initial road sub-semantic element meets the filtering rule, and the fused initial road sub-semantic element is determined as the initial road semantic element corresponding to the candidate road semantic element group. Some embodiments of the present specification can remove the influence of noise in the candidate road semantic element group through rule filtering to improve the accuracy of the initial road semantic element.

[0089] In some embodiments, similarly, for each candidate road semantic element group, the extraction module 220 may fuse the confidences of the initial road sub-semantic elements in the candidate road semantic element group to determine the confidence of the initial road semantic element corresponding to the candidate road semantic element group.

[0090] Step 440 : determining an initial road semantic element set based on the multiple initial road semantic elements.

[0091] The extraction module 220 may group multiple initial road semantic elements together to determine an initial road semantic element set.

[0092] Some embodiments of the present specification can process multiple original data sets separately to simply reduce the amount of calculation for each calculation, improve calculation efficiency, avoid the influence between similar road elements, and quickly and accurately determine the initial road semantic element set of the target area.

[0093] Figure 5 is an exemplary flow chart of determining a target road semantic element set of a target area according to some embodiments of this specification. In some embodiments, process 500 may be executed by optimization module 240. Figure 5 As shown, process 500 includes the following steps:

[0094] Step 510: Determine a first road structure change point based on a reference road element set.

[0095] The first road structure change point may refer to a point at which the road structure corresponding to the reference road element set characterizes a change.

[0096] The optimization module 240 can analyze and process each reference road element in the reference road element set to determine the first road structure change point. For example, when a lane line intersects with another lane line, it can be considered that the road structure has changed at the intersection. Correspondingly, the optimization module 240 can determine the intersection position of the lane line and the other lane line as the first road structure change point. For another example, the optimization module 240 finds through analysis and processing that a lane line overlaps with the position of a traffic light, that is, a traffic light is set on the road corresponding to the lane line, so the optimization module 240 can determine the position of the traffic light as the first road structure change point.

[0097] Step 520: Determine a plurality of first topological blocks according to the first road structure change point.

[0098] The first topological block may refer to a topological structure corresponding to an area between two adjacent first road structure change points in the target area represented by the reference road element set.

[0099] In some embodiments, the optimization module 240 can group the reference road element sets according to the first road structure change point, perform geometric structure topology expression modeling abstraction on each group of reference road element sets, and obtain a corresponding geometric topology map. Fig. 8A The reference road element sets corresponding to the reference map shown in the figure are grouped, wherein one set of reference road element sets is related to road 810a, and another set of reference road element sets is related to roads 820a and 830a. The optimization module 240 can perform geometric structure topological modeling abstract expression on the two sets of reference road element sets to obtain the following: Figure 8B The first topology block 840 and the first topology block 850 are shown, wherein Fig. 8A The roads 810a, 820a and 830a in FIG. 8 correspond to Figure 8B The topological element 810b, the topological element road 820b and the topological element road 830b in it.

[0100] Step 530: Determine a second road structure change point according to the initial road semantic element set.

[0101] The second road structure change point may refer to a point at which the road structure corresponding to the initial road semantic element set changes.

[0102] For more information about determining the second road structure change point according to the initial road semantic element set, please refer to the above description of determining multiple first topological blocks according to the first road structure change point.

[0103] Step 540: Determine a plurality of second topological blocks according to the second road structure change point.

[0104] The second topological block may refer to a topological structure corresponding to an area between two adjacent second road structure change points in the target area represented by the initial road semantic element set.

[0105] For more information about determining the plurality of second topological blocks according to the second road structure change point, reference may be made to the aforementioned description of determining the first road structure change point based on the reference road element set.

[0106] Step 550: Determine a first topological block that matches the second topological block.

[0107] In some embodiments, for each second topological block, the optimization module 240 can filter out a first topological block within a preset range corresponding to the position from multiple first topological blocks based on the position corresponding to the second topological block (for example, a first topological block 5 meters near the position), where the aforementioned preset range is related to the accuracy of the reference map.

[0108] In some embodiments, for each second topological block, the optimization module 240 may compare the second topological block and the filtered first topological block based on particle filtering or other feasible algorithms, so as to determine the first topological block corresponding to the second topological block.

[0109] Step 560: Optimize the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block to obtain the target road semantic element of the second topological block.

[0110] In some embodiments, for each second topological block, the optimization module 240 may determine whether the initial road semantic element corresponding to the second topological block satisfies the optimization condition. The aforementioned optimization condition may refer to the condition that the initial road semantic element corresponding to the second topological block needs to be optimized.

[0111] The aforementioned optimization conditions may include a first optimization condition and / or a second optimization condition.

[0112] The first optimization condition may include that the initial road semantic element corresponding to the second topological block is incomplete or the confidence is less than a preset confidence threshold. For example, when the shape of the lane line in the initial road semantic element corresponding to the second topological block is incomplete, the initial road semantic element corresponding to the second topological block satisfies the first optimization condition. For another example, when the confidence corresponding to the lane marking in the initial road semantic element corresponding to the second topological block is 0.6, and the preset confidence threshold is 0.8, the initial road semantic element corresponding to the second topological block satisfies the first optimization condition.

[0113] The second optimization condition may include a mismatch between an initial road semantic element corresponding to the second topological block and a reference road element corresponding to the first topological block that matches the second topological block. For example, when the type of lane marking in the initial road semantic element corresponding to the second topological block is straight ahead only, but the type of lane marking in the reference road element corresponding to the first topological block that matches the second topological block is straight ahead and left turn, the initial road semantic element corresponding to the second topological block satisfies the second optimization condition. For another example, when the reference road element corresponding to the first topological block that matches the second topological block includes a zebra crossing, but the initial road semantic element corresponding to the second topological block does not contain a zebra crossing, the initial road semantic element corresponding to the second topological block satisfies the second optimization condition.

[0114] In some embodiments, when the preset optimization conditions are met, the optimization module 240 can optimize the initial road semantic elements corresponding to the second topological block based on the reference road elements corresponding to the first topological block matching the second topological block through multiple methods to obtain the target road semantic elements of the second topological block.

[0115] In some embodiments, when the initial road semantic element corresponding to the second topological block meets the first optimization condition, the optimization module 240 may process the initial road semantic element corresponding to the second topological block and the reference road element corresponding to the first topological block matching the second topological block based on the optimization model to determine the target road semantic element corresponding to the second topological block. The aforementioned optimization model may be a machine learning model, such as a deep learning model.

[0116] The optimization module 240 may input the initial road semantic element corresponding to the second topological block and the reference road element corresponding to the first topological block matching the second topological block into the optimization model, and the output of the optimization model is the target road semantic element corresponding to the second topological block.

[0117] The optimization model can be obtained through training, and the training samples may include the sample initial road semantic elements corresponding to the sample second topological block and the sample reference road elements corresponding to the sample first topological block, wherein the aforementioned sample first topological block matches the sample second topological block, and the training labels may include the sample target road semantic elements obtained by manually annotating the area corresponding to the sample second topological block. In some embodiments, the input of the optimization model may also include the confidence of the initial road semantic elements corresponding to the second topological block, and correspondingly, when training the optimization model, the training samples may also include the sample confidence of the sample initial road semantic elements corresponding to the sample second topological block.

[0118] Some embodiments of the present specification can more accurately determine the target road semantic elements corresponding to the second topological block by processing the initial road semantic elements corresponding to the second topological block and the reference road elements corresponding to the first topological block matching the second topological block through an optimization model, thereby avoiding the cost of manual processing.

[0119] In some embodiments, when the initial road semantic element corresponding to the second topological block satisfies the first optimization condition, the optimization module 240 may also determine the road semantic element to be optimized corresponding to the second topological block based on the initial road semantic element corresponding to the second topological block. The road semantic element to be optimized may refer to the road semantic element that needs to be optimized. For example, the optimization module 240 may determine the road semantic element in the initial road semantic element whose confidence is less than a preset confidence threshold as the road semantic element to be optimized. For another example, the optimization module 240 may determine the incomplete road semantic element in the initial road semantic element as the road semantic element to be optimized. For another example, the optimization module 240 may determine the missing road semantic element in the initial road semantic element as the road semantic element to be optimized.

[0120] In some embodiments, the optimization module 240 can optimize the road semantic element to be optimized based on the reference road element corresponding to the first topological block matching the second topological block, and determine the target road semantic element corresponding to the second topological block. When the initial road semantic element corresponding to the second topological block meets the first optimization condition, the optimization module 240 can determine the reference road element corresponding to the road semantic element to be optimized from the corresponding reference road element in the first topological block based on the road semantic element to be optimized, and the optimization module 240 can convert the reference road element corresponding to the road semantic element to be optimized into an optimization rule, optimize the road semantic element to be optimized based on the optimization rule, and determine the target road semantic element corresponding to the second topological block. The optimization rule can be a rule for optimizing the road semantic element to be optimized. The correspondence between the optimization rule and the reference road element can be determined based on a pre-set setting. It can be understood that due to the actual rule, the constraint conditions of the corresponding area can be determined based on the reference road element, thereby determining the corresponding optimization rule. For example, when the reference road element includes a speed limit sign with an attribute feature of a speed limit of 80M / H to 120M / H, that is, it indicates that the area is on a highway, the corresponding optimization rule may include that there are no traffic lights, zebra crossings, etc. in the corresponding area.

[0121] Exemplarily, when the road semantic element to be optimized is a zebra crossing with low confidence, the optimization module 240 can determine the speed limit sign of 80M / H to 120M / H in the reference road element of the corresponding first topological block as the reference road element corresponding to the road semantic element to be optimized, and based on the aforementioned reference road element, determine the optimization rule of the road semantic element to be optimized (for example, there are no traffic lights or zebra crossings in the area), thereby determining that the target road semantic element corresponding to the partial area does not include a zebra crossing.

[0122] In some embodiments of the present specification, road semantic elements to be optimized can be screened out from initial road semantic elements. By optimizing the road semantic elements to be optimized, the accuracy of the target road semantic elements can be improved, avoiding the waste of computing resources caused by optimizing all initial road semantic elements.

[0123] In some embodiments, the optimization module 240 may also determine the road semantic element to be optimized corresponding to the second topological block based on the initial road semantic element corresponding to the second topological block; determine whether the road semantic element to be optimized needs to meet a preset accuracy requirement; in response to determining that the road semantic element to be optimized does not need to meet the preset accuracy requirement, determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block; and determine the reference road element corresponding to the road semantic element to be optimized as the target road semantic element. In some embodiments, the optimization module 240 may also determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block in response to the road semantic element to be optimized needing to meet the preset accuracy requirement; generate a labeling prompt for the road semantic element to be optimized based on the reference road element corresponding to the road semantic element to be optimized; and determine the target road semantic element based on the road semantic element to be optimized and the labeling prompt. For more information about the above two embodiments, please refer to Figure 6 and its related description.

[0124] Step 570: Determine a target road semantic element set of the target area based on the target road semantic elements corresponding to the plurality of second topological blocks.

[0125] It is understandable that, as mentioned above, the reference map may contain errors. If the reference road element set and the initial road semantic element set are matched directly based on the corresponding position relationship, errors may occur. Some embodiments of the present specification can improve the matching accuracy and avoid errors by processing the reference road elements into multiple first topological blocks, processing the initial road semantic element set into multiple second topological blocks, and then finding the first topological block corresponding to the second topological block. Some embodiments of the present specification can also optimize the initial road semantic elements corresponding to the second topological block based on the reference road elements corresponding to the first topological block that matches the second topological block, determine the target road semantic elements corresponding to the second topological block, and with the assistance of the reference map, realize automatic extraction of road semantic elements to adapt to complex road conditions.

[0126] Figure 6 FIG. 6 is another exemplary flow chart of determining a target road semantic element according to some embodiments of the present specification. In some embodiments, process 600 may be executed by the optimization module 240 .

[0127] In some embodiments, when the initial road semantic element corresponding to the second topological block meets the second optimization condition, the optimization module 240 can optimize the initial road semantic element corresponding to the second topological block by executing the process 600 to obtain the target road semantic element of the second topological block. Figure 6 As shown, process 600 includes the following steps:

[0128] Step 610: Determine the road semantic element to be optimized corresponding to the second topological block.

[0129] For more information about the road semantic elements to be optimized and determining the road semantic elements to be optimized corresponding to the second topological block, see Figure 5 and its related description.

[0130] Step 620: Determine whether the road semantic element to be optimized needs to meet a preset accuracy requirement.

[0131] The preset accuracy requirement may refer to the pre-set limit requirement on the accuracy of the target road semantic element. It is understandable that some road elements do not require high accuracy, and the accuracy corresponding to the reference road elements extracted based on the reference map is sufficient, and there is no need to meet the preset accuracy requirement, such as no-parking sections and speed limit signs; some road elements have high accuracy requirements and need to meet the preset accuracy requirements, such as lane lines and traffic lights.

[0132] The optimization module 240 may determine the road element corresponding to the road semantic element to be optimized, and determine whether the road semantic element to be optimized needs to meet the preset accuracy requirement based on the correspondence between the preset road elements and the preset accuracy requirement.

[0133] Step 630 : In response to determining that the road semantic element to be optimized does not need to meet the preset accuracy requirement, determine a reference road element corresponding to the road semantic element to be optimized from reference road elements corresponding to the first topological block matching the second topological block.

[0134] For example, when the road semantic element to be optimized is a no-parking section, the optimization module 240 may determine a reference road element related to the no-parking section from the reference road elements corresponding to the first topological block that matches the second topological block.

[0135] Step 640: determine the reference road element corresponding to the road semantic element to be optimized as the target road semantic element.

[0136] In some embodiments of the present specification, an evaluation is performed on whether the road semantic element to be optimized needs to meet preset accuracy requirements, so as to determine how to optimize the road semantic element to be optimized based on the reference road element, thereby avoiding errors caused by inconsistent accuracy requirements during direct replacement.

[0137] In some embodiments, in response to determining that the road semantic element to be optimized needs to meet the preset accuracy requirement, the optimization module 240 can determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block; based on the reference road element corresponding to the road semantic element to be optimized, generate a labeling prompt for the road semantic element to be optimized. The aforementioned labeling prompts prompt information of the road semantic element to be optimized. The labeling prompt may include the reference road element corresponding to the road semantic element to be optimized and the reference mark. The reference road element can provide a reference for the road semantic element to be optimized. The reference mark indicates that the accuracy of the reference road element is low and can only be used as a reference. For example, when the road semantic element to be optimized is a missing traffic light, the optimization module 240 can generate a labeling prompt corresponding to the road semantic element to be optimized (e.g., missing road semantic element: traffic light and its type, position, shape and other attribute features) based on the reference road element corresponding to the road semantic element to be optimized (i.e., the traffic light and its attribute features in the corresponding reference road element).

[0138] In some embodiments, the optimization module 240 may determine the target road semantic element based on the road semantic element to be optimized and the annotation prompt. The optimization module 240 may determine the road semantic element to be optimized with the annotation prompt as the target road semantic element.

[0139] In some embodiments of the present specification, when the road semantic elements to be optimized need to meet preset accuracy requirements, the road semantic elements to be optimized can be prompted by marking prompts, so as to provide references for the road semantic elements to be optimized based on the reference road elements without reducing the accuracy of the target road semantic elements to the maximum extent, thereby facilitating the subsequent development of other work.

[0140] Some embodiments of the present specification also provide a road semantic element generation device, including a processor, wherein the processor is used to execute the road semantic element generation method as described in any of the aforementioned embodiments of the present specification.

[0141] Some embodiments of the present specification also provide a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the road semantic element generation method as described in any of the aforementioned embodiments of the present specification.

[0142] The solutions described in this specification and in the examples, if they involve the processing of personal information, will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or being necessary for the performance of a contract, etc.), and will only be processed within the scope of regulations or agreements. If a user refuses to process personal information other than the necessary information for basic functions, it will not affect the user's use of basic functions.

[0143] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0144] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0145] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0146] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0147] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0148] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0149] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A method for generating road semantic elements, characterized in that: include: Obtain original road data of the target area; Extracting an initial road semantic element set of the target area from the original road data, wherein the initial road semantic element set includes a plurality of initial road semantic elements; Acquire a reference road element set related to the target area, wherein the reference road element set includes a plurality of reference road elements; The initial road semantic element set is optimized according to the reference road element set to obtain a target road semantic element set of the target area, wherein the target road semantic element set includes multiple target road semantic elements, and the accuracy of the multiple target road semantic elements is greater than the accuracy of the multiple reference road elements.

2. The method according to claim 1, characterized in that The original road data of the target area is divided into a plurality of original data sets, each of which includes the original road data obtained by detecting the target area at a detection time; The extracting of the initial road semantic element set from the original road data comprises: For each of the original data sets, the original road data in the original data set is processed based on a semantic detection model to determine an initial road sub-semantic element corresponding to the original data set, wherein the semantic detection model is a machine learning model; Based on the geometric relationships and attribute features of the initial road sub-semantic elements corresponding to the multiple original data sets, the initial road sub-semantic elements corresponding to the multiple original data sets are divided into multiple candidate road semantic element groups; Merging the initial road sub-semantic elements in each of the candidate road semantic element groups to determine the multiple initial road semantic elements; Based on the multiple initial road semantic elements, the initial road semantic element set is determined.

3. The method according to claim 1, characterized in that The optimizing the initial road semantic element set according to the reference road element set to obtain the target road semantic element set of the target area comprises: Based on the reference road element set, determining a first road structure change point; Determining a plurality of first topological blocks according to the first road structure change point; Determining a second road structure change point according to the initial road semantic element set; determining a plurality of second topological blocks according to the second road structure change point; For each of the second topological blocks, Determining the first topological block that matches the second topological block; Optimizing the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block, to obtain a target road semantic element of the second topological block; The target road semantic element set of the target area is determined based on the target road semantic elements corresponding to the multiple second topological blocks.

4. The method according to claim 3, characterized in that The optimizing the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block to obtain the target road semantic element of the second topological block includes: The initial road semantic element corresponding to the second topological block and the reference road element corresponding to the first topological block matching the second topological block are processed based on the optimization model to determine the target road semantic element corresponding to the second topological block.

5. The method according to claim 3, characterized in that The optimizing the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block to obtain the target road semantic element of the second topological block includes: Determining, based on the initial road semantic element corresponding to the second topological block, a road semantic element to be optimized corresponding to the second topological block; The road semantic element to be optimized is optimized based on the reference road element corresponding to the first topological block matching the second topological block, and the target road semantic element corresponding to the second topological block is determined.

6. The method according to claim 3, characterized in that The optimizing the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block to obtain the target road semantic element of the second topological block includes: Determining, based on the initial road semantic element corresponding to the second topological block, a road semantic element to be optimized corresponding to the second topological block; Determining whether the road semantic element to be optimized needs to meet a preset accuracy requirement; In response to determining that the road semantic element to be optimized does not need to meet the preset accuracy requirement, Determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block; The reference road element corresponding to the road semantic element to be optimized is determined as the target road semantic element.

7. The method according to claim 6, characterized in that The optimizing the initial road semantic element corresponding to the second topological block based on the reference road element corresponding to the first topological block matching the second topological block to obtain the target road semantic element of the second topological block includes: In response to the road semantic element to be optimized needing to meet the preset accuracy requirement, Determine the reference road element corresponding to the road semantic element to be optimized from the reference road elements corresponding to the first topological block matching the second topological block; Based on the reference road element corresponding to the road semantic element to be optimized, generating a labeling prompt of the road semantic element to be optimized; The target road semantic element is determined based on the road semantic element to be optimized and the annotation prompt.

8. A road semantic element generation system, characterized in that: include: A first acquisition module is used to acquire original road data of a target area; An extraction module, used for extracting an initial road semantic element set of the target area from the original road data, wherein the initial road semantic element set includes a plurality of initial road semantic elements; A second acquisition module, configured to acquire a reference road element set related to the target area, wherein the reference road element set includes a plurality of reference road elements; An optimization module is used to optimize the initial road semantic element set according to the reference road element set to obtain a target road semantic element set of the target area, wherein the target road semantic element set includes multiple target road semantic elements, and the accuracy of the multiple target road semantic elements is greater than the accuracy of the multiple reference road elements.

9. A road semantic element generation device, comprising a processor, characterized in that: The processor is used to execute the road semantic element generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the road semantic element generation method according to any one of claims 1 to 7.

Images