Map generation method and device, vehicle and storage medium

Abstract

The invention relates to a map generation method and device, a vehicle and a storage medium. An environment image of the current position of the vehicle can be acquired; determining a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used for describing geographical position information of at least one target object in a first preset area around the vehicle; performing matching fusion on the semantic map and the navigation map of the vehicle through a Transform model obtained through pre-training to obtain map fusion data; and adding the geographic position information of at least one target object into the navigation map according to the map fusion data, and then generating a target map.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to a method, device, vehicle and storage medium for generating a map. Background Art

[0002] With the progress of the times and the development of technology, self-driving cars will become a reality, and the arrival of self-driving cars will provide great convenience to our lives. In self-driving technology, high-precision maps have become an important part of it. Traditional in-vehicle maps usually use satellite images or satellite data to generate navigation maps, which have certain limitations in accuracy and cannot meet the needs of autonomous driving systems for high-precision maps. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides a method, device, vehicle and storage medium for generating a map.

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided a method for generating a map, comprising:

[0005] Get the environment image of the vehicle's current location;

[0006] Determine a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle;

[0007] After matching and fusing the semantic map and the navigation map of the vehicle through a pre-trained Transformer model, map fusion data is obtained;

[0008] After adding geographical location information of at least one target object to the navigation map according to the map fusion data, a target map is generated.

[0009] Optionally, the Transformer model includes a cross attention mechanism, and after the pre-trained Transformer model matches and fuses the semantic map and the navigation map of the vehicle, the map fusion data obtained includes:

[0010] Acquire navigation map data corresponding to the first preset area in the navigation map;

[0011] Performing vectorization processing on the navigation map data and the semantic map respectively to obtain a first vector corresponding to the navigation map data and a second vector corresponding to the semantic map;

[0012] According to the first vector and the second vector, the semantic map and the navigation map are matched and fused through the cross attention mechanism to obtain the map fusion data.

[0013] Optionally, the obtaining of the map fusion data after matching and fusing the semantic map and the navigation map through the cross attention mechanism according to the first vector and the second vector includes:

[0014] For each element set corresponding to the second vector, performing attention calculation through the cross attention mechanism according to the vector elements in the element set and the first vector, to obtain fused data of the target object corresponding to the element set, wherein the fused data represents the world coordinates of the target object, the element set includes at least one vector element, and the element set corresponds to the target object one-to-one;

[0015] The navigation map data and the fused data respectively corresponding to each target object in the at least one target object are used as the map fusion data.

[0016] Optionally, the method further comprises:

[0017] Acquire high-precision map cache data in a second preset area corresponding to the current location, the first preset area is within the second preset area, and the second preset area is larger than the first preset area;

[0018] The target map is generated according to the high-precision map cache data and the map fusion data.

[0019] Optionally, the target map includes a first map corresponding to the first preset area and a second map corresponding to a difference area, and the difference area is an area corresponding to a difference between the second preset area and the first preset area;

[0020] The generating the target map according to the high-precision map cache data and the map fusion data comprises:

[0021] generating the first map according to the map fusion data;

[0022] Acquire target map cache data corresponding to the difference set area from the high-precision map cache data;

[0023] The second map is generated according to the target map cache data.

[0024] Optionally, the environment image includes a bird's-eye view BEV; and determining the semantic map corresponding to the current position according to the environment image includes:

[0025] After performing BEV perception on the bird's-eye view, determining an identification of each of the target objects and a relative position of each of the target objects to the vehicle;

[0026] For each of the target objects, the semantic map is determined according to the identifier and the relative position of the target object.

[0027] Optionally, the method further comprises:

[0028] After the target map is generated, the high-precision map data of the target map corresponding to the current position is stored in a preset map database.

[0029] According to a second aspect of an embodiment of the present disclosure, there is provided a device for generating a map, comprising:

[0030] An acquisition module is configured to acquire an environmental image of a current position of the vehicle;

[0031] A determination module, configured to determine a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle;

[0032] A fusion module is configured to match and fuse the semantic map and the navigation map of the vehicle through a pre-trained Transformer model to obtain map fusion data;

[0033] The map generation module is configured to generate a target map after adding at least one geographical location information of the target object to the navigation map according to the map fusion data.

[0034] According to a third aspect of an embodiment of the present disclosure, a vehicle is provided, comprising:

[0035] processor;

[0036] a memory for storing processor-executable instructions;

[0037] Wherein, the processor is configured to: execute the steps of the method for generating a map described in the first aspect of the present disclosure.

[0038] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the program instructions are executed by a processor, the steps of the method for generating a map provided in the first aspect of the present disclosure are implemented.

[0039] The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: determining a semantic map corresponding to the current position of the vehicle based on an environmental image, since the semantic map is used to describe the geographic location information of at least one target object in the first preset area around the vehicle, after generating a target map based on the semantic map and the navigation map of the vehicle, compared with the navigation map, the target map is added with the geographic location information of at least one target object (such as lane lines, crosswalks, etc.), and the accuracy of the target map is higher. In addition, the complex operation of generating a high-precision map based on manually collected data is avoided. At the same time, since the environmental image is an environmental image of the current position of the vehicle acquired in real time, the semantic map generated based on the environmental image can reflect the latest changes of the target objects in the environment around the vehicle, thereby ensuring that the generated target map is the latest map corresponding to the current changes in geographic information, thereby improving the accuracy of the generated target map.

[0040] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0042] Figure 1 The present invention is a flowchart of a method for generating a map according to an exemplary embodiment.

[0043] Figure 2 is based on Figure 1 The illustrated embodiment is a flow chart of another method for generating a map.

[0044] Figure 3 The present invention is a schematic diagram of a process of matching and fusing a semantic map with a navigation map according to an exemplary embodiment.

[0045] Figure 4 is based on Figure 1 The illustrated embodiment shows a flow chart of a method for generating a map.

[0046] Figure 5 is based on Figure 1 The illustrated embodiment shows a flow chart of a method for generating a map.

[0047] Figure 6 It is a block diagram of a device for generating a map according to an exemplary embodiment.

[0048] Figure 7 is based on Figure 6 The illustrated embodiment shows a block diagram of a device for generating a map.

[0049] Figure 8 is a block diagram of a vehicle according to an exemplary embodiment. DETAILED DESCRIPTION

[0050] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0051] It should be noted that all actions to obtain signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the location and with the authorization given by the owner of the corresponding device.

[0052] The present disclosure is mainly used in scenarios where vehicles perform autonomous driving based on high-precision maps. Currently, high-precision maps mainly rely on manual collection, which is costly and inefficient. On the one hand, high-precision maps need to collect a large amount of data, including information such as roads, buildings, traffic signs, traffic lights, etc., which requires a lot of manpower, material resources and financial resources; on the other hand, high-precision maps need to be constantly updated to reflect the changes and development of the city. Obviously, the existing high-precision maps generated by manual collection cannot reflect the changes in geographic information in real time.

[0053] In order to solve the above problems, the present invention provides a method, device, vehicle and storage medium for generating a map. The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings.

[0054] Figure 1 FIG. 1 is a flow chart of a method for generating a map according to an exemplary embodiment, and the method can be applied to a vehicle. Figure 1 As shown, the method includes the following steps.

[0055] In step S11 , an environment image of the current position of the vehicle is acquired.

[0056] The environment image may include a BEV (Bird's-Eye View).

[0057] In step S12, a semantic map corresponding to the current position is determined according to the environmental image, where the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle.

[0058] Among them, the first preset area may include an area centered on the vehicle and within a preset distance range around the vehicle, and may also include an area within a preset distance range from the vehicle in the direction of vehicle travel. The present disclosure does not make any specific limitations on the first preset area, and it can be set arbitrarily according to actual needs.

[0059] The target object usually refers to geographic feature information that does not exist in the navigation map. For example, the target object may include lane lines, crosswalks, road markings, etc. In existing navigation maps generated based on satellite images or satellite data, geographic location information such as lane lines, crosswalks, road markings, etc. is not marked in the navigation map. For autonomous driving vehicles, it is necessary to perform environmental perception of the location information of lane lines, crosswalks, road markings, etc. on the road where the vehicle is located. Therefore, existing navigation maps cannot meet the needs of autonomous driving vehicles for high-precision maps.

[0060] The geographic location information may include relative location information of each target object relative to the vehicle. That is, in the present disclosure, the semantic map may mark relative location information of at least one target object in a first preset area around the vehicle relative to the vehicle.

[0061] In a possible implementation of this step, after performing BEV perception on the bird's-eye view, the identification of each target object and the relative position of each target object to the vehicle can be determined; for each target object, the semantic map can be determined based on the identification and the relative position of the target object.

[0062] The identifier of the target object may be, for example, a specific category or name of the target object (such as a lane line, a crosswalk, or a road marking line).

[0063] That is to say, after the bird's-eye view is input into a pre-trained BEV perception model, the BEV perception model can output the name of each target object in the first preset area and the position information of each target object relative to the vehicle.

[0064] In step S13, the semantic map and the navigation map of the vehicle are matched and fused by a pre-trained Transformer model to obtain map fusion data.

[0065] The map fusion data includes the geographical location corresponding to each target object, and the geographical location may include the world coordinates of the corresponding target object, for example. The Transformer model includes a cross-attention mechanism. The cross-attention mechanism refers to the attention calculation of a certain position in a sequence and all positions in another sequence. In the present disclosure, through the cross-attention mechanism, the Transformer model can effectively use the information of the semantic map and the navigation map to calculate the attention in the process of generating map fusion data, thereby improving the accuracy of the generated map data.

[0066] In step S14, after adding the geographical location information of at least one target object to the navigation map according to the map fusion data, a target map is generated.

[0067] The target map refers to a high-precision map (HD map) in the first preset area generated based on the semantic map in the first preset area and the navigation map. The HD map may include map elements such as lane lines, road markings, traffic signs and obstacles, and its map accuracy can be down to the centimeter level.

[0068] In the process of generating the target map, the target map can be generated by adding the geographical location information of target objects such as lane lines, crosswalks, road marking lines, etc. in the first preset area around the vehicle to the navigation map according to the semantic map. In this way, the target map can significantly improve the map accuracy compared with the navigation map.

[0069] In one possible implementation, the world coordinates of each target object in the first preset area can be obtained based on the map fusion data, and then each target object in the first preset area can be added to the navigation map based on the world coordinates, thereby generating a high-precision map of the target.

[0070] The Transformer model is used to process the semantic map and the navigation map data of the vehicle map software in real time to generate a higher-precision target map. When the vehicle navigates based on the target map, the accuracy and safety of vehicle navigation can be improved.

[0071] The above method is used to determine the semantic map corresponding to the current position of the vehicle based on the environmental image. Since the semantic map is used to describe the geographic location information of at least one target object in the first preset area around the vehicle, after the target map is generated based on the semantic map and the navigation map of the vehicle, compared with the navigation map, the target map is added with the geographic location information of at least one target object (such as lane lines, crosswalks, etc.), and the accuracy of the target map is higher. In addition, the complex operation of generating high-precision maps based on manually collected data is avoided. At the same time, since the environmental image is an environmental image of the current position of the vehicle obtained in real time, the semantic map generated based on the environmental image can reflect the latest changes of the target objects in the environment around the vehicle, thereby ensuring that the generated target map is the latest map corresponding to the current changes in geographic information, thereby improving the accuracy of the generated target map.

[0072] Figure 2 is based on Figure 1 The embodiment shown is a flowchart of another method for generating a map, such as Figure 2 As shown, step S13 includes the following sub-steps:

[0073] In step S131, navigation map data corresponding to the first preset area in the navigation map is obtained.

[0074] As mentioned above, the first preset area can be set arbitrarily according to actual needs. It should be noted that the larger the first preset area, the larger the area where the vehicle needs to perform BEV perception of the environmental image based on the BEV perception model, and the longer the model calculation time required, which will affect the real-time generation of the high-precision map. Therefore, when setting the first preset area, the real-time requirements of the current business scenario for the generation of the high-precision map must also be considered.

[0075] In this step, only the navigation map data of the navigation map within the first preset area around the vehicle can be obtained, so that a high-precision map within the first preset area can be generated based only on the navigation map data within the first preset area and the semantic map within the first preset area. In addition, the navigation map can be an in-vehicle map software, and the navigation map data can include data such as the name and type of the geographical feature corresponding to each position on the navigation map.

[0076] In step S132, vectorization processing is performed on the navigation map data and the semantic map respectively to obtain a first vector corresponding to the navigation map data and a second vector corresponding to the semantic map.

[0077] In this step, the navigation map data and the map data of the semantic map may be vectorized based on a vectorization method provided in relevant literature to obtain the first vector and the second vector.

[0078] In step S133, according to the first vector and the second vector, the semantic map and the navigation map are matched and fused through the cross-attention mechanism to obtain the map fusion data.

[0079] The map fusion data may include navigation map data corresponding to the first preset area and fusion data corresponding to each target object in the first preset area. The fusion data may include, for example, the world coordinates of the target object.

[0080] In this step, for each element set corresponding to the second vector, attention calculation can be performed through the cross-attention mechanism according to the vector elements in the element set and the first vector to obtain fusion data of the target object corresponding to the element set, wherein the fusion data represents the world coordinates of the target object, the element set includes at least one vector element, and the element set corresponds one-to-one to the target object; the navigation map data and the fusion data corresponding to each target object in the at least one target object are used as the map fusion data.

[0081] For example, Figure 3 is a schematic diagram of a process of matching and fusing a semantic map with a navigation map according to an exemplary embodiment. Figure 3 As shown, the second vector corresponding to the semantic map based on the BEV perception output is input into the cross-attention mechanism model (i.e., the cross-attention model), and the first vector corresponding to the navigation map in the first preset area is also input into the cross-attention model, as shown in FIG. Figure 3 As shown, a circle represents an element set (including at least one vector element) in the second vector, and an element set corresponds to a target object in the first preset area one by one. Thus, for each target object in the first preset area, at least one vector element is used to represent the relative position information between the target object and the vehicle. Thus, Figure 3 The cross-attention model shown in the figure can calculate the attention of the at least one vector element and the first vector (the vector representation of the navigation map data in the first preset area) to obtain the fused data corresponding to the target object, and the fused data represents the world coordinates of the target object. Figure 3 The matching and fusion method shown can obtain the world coordinates of each target object. The above examples are only for illustration and are not limited in the present disclosure.

[0082] As mentioned above, considering that there will be different requirements for the real-time generation of high-precision maps in actual business scenarios, therefore, in order to ensure that high-precision maps can be generated quickly, the first preset area usually needs to be set smaller. In another possible application scenario, the high-precision map that needs to be generated must have a larger area range as much as possible while ensuring that it can be generated in real time. In this case, the high-precision map cache data in the pre-generated second preset area can be obtained, and the high-precision map cache data in the second preset area is fused with the map fusion data in the latest generated first preset area to generate a high-precision map with a larger area range (i.e., the second preset area). Among them, the first preset area is located in the second preset area, and the second preset area is larger than the first preset area.

[0083] For example, assuming that the first preset area is a circular area centered on the vehicle and within a range of 5 meters around the vehicle, the second preset area can be a circular area centered on the vehicle and within a range of 10 meters around the vehicle. It can be understood that the first preset area can be used as an inner circle and the second preset area can be used as an outer circle, and the centers of the two circles are both the vehicle.

[0084] Figure 4 is based on Figure 1 The embodiment shown is a flowchart of a method for generating a map, such as Figure 4 As shown, the method further comprises the following steps:

[0085] In step S15, high-precision map cache data in a second preset area corresponding to the current position is obtained, the first preset area is located in the second preset area, and the second preset area is larger than the first preset area.

[0086] The current location may be the GPS longitude and latitude information of the vehicle. In this way, the high-precision map cache data in the second preset area can be obtained from the map database based on the current GPS longitude and latitude information of the vehicle.

[0087] It should be noted that the high-precision map cache data in the second preset area may be map data stored in the map database after the vehicle generated a high-precision map last time.

[0088] In this way, during the execution of step S14, the target map can be generated based on the high-precision map cache data and the map fusion data.

[0089] The target map includes a first map corresponding to the first preset area and a second map corresponding to a difference area, and the difference area is an area corresponding to the difference between the second preset area and the first preset area.

[0090] In this way, in the process of generating the target map based on the high-precision map cache data and the map fusion data, the first map can be generated based on the map fusion data; the target map cache data corresponding to the difference set area is obtained from the high-precision map cache data; and the second map is generated based on the target map cache data.

[0091] For example, assuming that the first preset area is a circular area centered on the vehicle and within a range of 5 meters around the vehicle, and the second preset area can be a circular area centered on the vehicle and within a range of 10 meters around the vehicle, the first preset area can be used as an inner circle, and the second preset area can be used as an outer circle, and the centers of both circles are the vehicle, then the difference area is a circular ring area composed of the inner circle and the outer circle, and the width of the circular ring area is 5 meters. That is to say, in the process of generating a target map based on high-precision map cache data and map fusion data, the first map corresponding to the inner circle area can be generated based on the map fusion data corresponding to the first preset area, and the second map corresponding to the circular ring area can be generated based on the target map cache data corresponding to the circular ring area. In this way, the first map and the second map are combined together as the target map. This is only an example, and the present disclosure is not limited to this.

[0092] In this way, the target map is generated based on the high-precision map cache data in the map database and the map fusion data within the first preset area generated in real time, which can realize the reuse of high-precision map information and take into account the changed parts of the map information, thereby realizing real-time updating of the map information.

[0093] Figure 5 is based on Figure 1 The embodiment shown is a flowchart of a method for generating a map, such as Figure 5 As shown, the method further comprises the following steps:

[0094] In step S16, after the target map is generated, the high-precision map data of the target map corresponding to the current position is stored in a preset map database.

[0095] That is to say, after the target map is generated this time, the high-precision map data of the newly generated target map can be stored, so that the current high-precision map data can be reused in the subsequent generation of high-precision maps.

[0096] Figure 6 is a block diagram of a device for generating a map according to an exemplary embodiment. Figure 6 As shown, the device comprises:

[0097] An acquisition module 601 is configured to acquire an environmental image of a current position of the vehicle;

[0098] A determination module 602 is configured to determine a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle;

[0099] A fusion module 603 is configured to match and fuse the semantic map and the navigation map of the vehicle through a pre-trained Transformer model to obtain map fusion data;

[0100] The map generation module 604 is configured to generate a target map after adding geographical location information of at least one target object to the navigation map according to the map fusion data.

[0101] Optionally, the Transformer model includes a cross-attention mechanism, and the fusion module 603 is configured to obtain navigation map data corresponding to the first preset area in the navigation map; vectorize the navigation map data and the semantic map respectively to obtain a first vector corresponding to the navigation map data and a second vector corresponding to the semantic map; and match and fuse the semantic map and the navigation map through the cross-attention mechanism according to the first vector and the second vector to obtain the map fusion data.

[0102] Optionally, the fusion module 603 is configured to perform attention calculation through the cross-attention mechanism for each element set corresponding to the second vector, based on the vector elements in the element set and the first vector, to obtain fusion data of the target object corresponding to the element set, wherein the fusion data represents the world coordinates of the target object, the element set includes at least one vector element, and the element set corresponds one-to-one to the target object; and the fusion data corresponding to the navigation map data and each target object in the at least one target object are used as the map fusion data.

[0103] Optionally, the map generation module 604 is configured to obtain high-precision map cache data within a second preset area corresponding to the current position, the first preset area is located within the second preset area, and the second preset area is larger than the first preset area; and generate the target map based on the high-precision map cache data and the map fusion data.

[0104] Optionally, the target map includes a first map corresponding to the first preset area and a second map corresponding to a difference area, and the difference area is an area corresponding to a difference between the second preset area and the first preset area;

[0105] The map generation module 604 is configured to generate the first map according to the map fusion data; obtain the target map cache data corresponding to the difference set area from the high-precision map cache data; and generate the second map according to the target map cache data.

[0106] Optionally, the environmental image includes a bird's-eye view BEV; the determination module 602 is configured to determine the identification of each target object and the relative position of each target object to the vehicle after performing BEV perception on the bird's-eye view; for each target object, determine the semantic map according to the identification and the relative position of the target object.

[0107] Optionally, Figure 7 is based on Figure 6 The block diagram of a device for generating a map shown in the embodiment shown in FIG. Figure 7 As shown, the device also includes:

[0108] The map data storage module 605 is configured to store the high-precision map data of the target map corresponding to the current position into a preset map database after generating the target map.

[0109] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0110] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method for generating a map provided by the present disclosure.

[0111] Figure 8 800 is a block diagram of a vehicle according to an exemplary embodiment. For example, vehicle 800 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles. Vehicle 800 may be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0112] Reference Figure 8 , the vehicle 800 may include various subsystems, for example, an infotainment system 810, a perception system 820, a decision control system 830, a drive system 840, and a computing platform 850. The vehicle 800 may also include more or fewer subsystems, and each subsystem may include multiple components. In addition, each subsystem and each component of the vehicle 800 may be interconnected by wire or wireless means.

[0113] In some embodiments, the infotainment system 810 may include a communication system, an entertainment system, and a navigation system, etc.

[0114] The perception system 820 may include several sensors for sensing information about the environment around the vehicle 800. For example, the perception system 820 may include a global positioning system (the global positioning system may be a GPS system, or a Beidou system or other positioning systems), an inertial measurement unit (IMU), a laser radar, a millimeter wave radar, an ultrasonic radar, and a camera.

[0115] The decision control system 830 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0116] The drive system 840 may include components that provide powered motion for the vehicle 800. In one embodiment, the drive system 840 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of an internal combustion engine, an electric motor, and an air compression engine. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0117] Some or all functions of the vehicle 800 are controlled by a computing platform 850. The computing platform 850 may include at least one processor 851 and a memory 852, and the processor 851 may execute instructions 853 stored in the memory 852.

[0118] The processor 851 may be any conventional processor, such as a commercially available CPU. The processor may also include a graphics processor (Graphic Process Unit, GPU), a field programmable gate array (Field Programmable Gate Array, FPGA), a system on chip (System on Chip, SOC), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC) or a combination thereof.

[0119] The memory 852 may be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0120] In addition to instructions 853 , memory 852 may also store data, such as road maps, route information, vehicle location, direction, speed, etc. The data stored in memory 852 may be used by computing platform 850 .

[0121] In the embodiment of the present disclosure, the processor 851 may execute instruction 853 to complete all or part of the steps of the above-mentioned method for generating a map.

[0122] In another exemplary embodiment, a computer program product is also provided. The computer program product includes a computer program executable by a programmable device, and the computer program has a code portion for executing the above-mentioned method for generating a map when executed by the programmable device.

[0123] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the present disclosure. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0124] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for generating a map, characterized in that: include: Get the environment image of the vehicle's current location; Determine a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle; After matching and fusing the semantic map and the navigation map of the vehicle through a pre-trained Transformer model, map fusion data is obtained; After adding geographical location information of at least one target object to the navigation map according to the map fusion data, a target map is generated.

2. The method according to claim 1, characterized in that: The Transformer model includes a cross attention mechanism. After the Transformer model obtained through pre-training matches and fuses the semantic map and the navigation map of the vehicle, the map fusion data obtained includes: Acquire navigation map data corresponding to the first preset area in the navigation map; Performing vectorization processing on the navigation map data and the semantic map respectively to obtain a first vector corresponding to the navigation map data and a second vector corresponding to the semantic map; According to the first vector and the second vector, the semantic map and the navigation map are matched and fused through the cross attention mechanism to obtain the map fusion data.

3. The method according to claim 2, characterized in that After matching and fusing the semantic map and the navigation map through the cross attention mechanism according to the first vector and the second vector, obtaining the map fusion data includes: For each element set corresponding to the second vector, performing attention calculation through the cross attention mechanism according to the vector elements in the element set and the first vector, to obtain fused data of the target object corresponding to the element set, wherein the fused data represents the world coordinates of the target object, the element set includes at least one vector element, and the element set corresponds to the target object one-to-one; The navigation map data and the fused data respectively corresponding to each target object in the at least one target object are used as the map fusion data.

4. The method according to claim 1, characterized in that The method further comprises: Acquire high-precision map cache data in a second preset area corresponding to the current location, the first preset area is within the second preset area, and the second preset area is larger than the first preset area; The target map is generated according to the high-precision map cache data and the map fusion data.

5. The method according to claim 4, characterized in that The target map includes a first map corresponding to the first preset area and a second map corresponding to a difference area, wherein the difference area is an area corresponding to a difference between the second preset area and the first preset area; The generating the target map according to the high-precision map cache data and the map fusion data comprises: generating the first map according to the map fusion data; Acquire target map cache data corresponding to the difference set area from the high-precision map cache data; The second map is generated according to the target map cache data.

6. The method according to claim 1, characterized in that The environment image includes a bird's-eye view BEV; and determining the semantic map corresponding to the current position according to the environment image includes: After performing BEV perception on the bird's-eye view, determining an identification of each of the target objects and a relative position of each of the target objects to the vehicle; For each of the target objects, the semantic map is determined according to the identifier and the relative position of the target object.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: After the target map is generated, the high-precision map data of the target map corresponding to the current position is stored in a preset map database.

8. A device for generating a map, characterized in that: include: An acquisition module is configured to acquire an environmental image of a current position of the vehicle; A determination module, configured to determine a semantic map corresponding to the current position according to the environment image, wherein the semantic map is used to describe geographic location information of at least one target object in a first preset area around the vehicle; A fusion module is configured to match and fuse the semantic map and the navigation map of the vehicle through a pre-trained Transformer model to obtain map fusion data; The map generation module is configured to generate a target map after adding at least one geographical location information of the target object to the navigation map according to the map fusion data.

9. A vehicle, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to: execute the steps of the method for generating a map as described in any one of claims 1-7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

Images