Map generation method, device, terminal device and storage medium

Abstract

The present invention discloses a map generation method, apparatus, terminal device, and storage medium. The method comprises: upon detecting a vehicle entering a parking lot and powering off, obtaining the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle was powered off; upon detecting the vehicle starting to leave the parking lot, constructing a second semantic map based on the vehicle's position and heading information at the time of power-off; upon detecting the restoration of the global positioning signal, collecting the second global positioning data, temporally associating the global positioning data with the semantic map; and splicing the first and second semantic maps based on the first and second global positioning data. This solution reduces map construction errors and improves the accuracy of parking lot map construction.

Description

Technical Field

[0001] The present invention relates to the field of map processing technology, and in particular to a map generation method, apparatus, terminal device and storage medium. Background Art

[0002] AVP (Automated Valet Parking) is an application of autonomous driving in parking scenarios. It provides fully automated valet parking, ultimately aiming to replace traditional manual valet parking, saving users significant parking time and resolving the problem of long parking queues during peak hours.

[0003] The AVP function requires a parking map. Current smart vehicles have sufficient positioning sensors and computing units to construct parking maps. However, most parking maps are indoors. When GNSS (Global Navigation Satellite System) signals are lost, the position estimation error during map construction increases, resulting in inaccurate absolute positions on the map. Summary of the Invention

[0004] The main purpose of the embodiments of the present invention is to provide a map generation method, apparatus, terminal device and storage medium, aiming to improve the accuracy of parking lot map construction.

[0005] To achieve the above object, an embodiment of the present invention provides a map generation method, which includes the following steps:

[0006] Upon detecting that a vehicle enters a parking lot and is powered off, obtaining the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, wherein the first global positioning data is temporally associated with the first semantic map, and a global positioning signal is lost within the parking lot;

[0007] When detecting that a vehicle starts to leave the parking lot, constructing a second semantic map based on the vehicle's position and heading information at the time of power-off;

[0008] After detecting that the global positioning signal has been restored, collecting second global positioning data, and temporally associating the second global positioning data with the second semantic map;

[0009] Based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map.

[0010] Optionally, the method further includes:

[0011] Based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected.

[0012] Optionally, before the step of obtaining the vehicle's position and heading information at the time of power-off, as well as the first semantic map constructed and the first global positioning data collected before the vehicle is powered off, upon detecting that the vehicle enters a parking lot and is powered off, the step further includes:

[0013] In a startup state before the vehicle enters the parking lot, continuously constructing a first semantic map of a preset area and collecting first global positioning data;

[0014] Associating the first global positioning data with corresponding track points on the first semantic map in time;

[0015] After detecting that the global positioning signal is lost, the first semantic map continues to be constructed until it is detected that the vehicle enters the parking lot and is powered off.

[0016] Optionally, before the step of splicing the first semantic map and the second semantic map to obtain a spliced ​​semantic map, the step further includes:

[0017] Determining whether the accuracy of the second global positioning data meets a preset standard;

[0018] If the accuracy of the second global positioning data reaches a preset standard, the step of splicing the first semantic map and the second semantic map is performed.

[0019] Optionally, the step of correcting the spliced ​​semantic map based on the first global positioning data, the second global positioning data, and preset constraint rules, and in combination with a preset map fusion optimization method, includes:

[0020] Obtaining an overlapping area between the first semantic map and the second semantic map in the spliced ​​semantic map, and using matching semantic elements in the overlapping area as mutual constraints; and / or using position and heading changes of associated elements in the first semantic map as constraints, and using position and heading changes of associated elements in the second semantic map as constraints;

[0021] Based on the first global positioning data and the second global positioning data, acquiring corresponding constrained positioning data;

[0022] Based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method, the positions and directions of all semantic elements in the spliced ​​semantic map are obtained;

[0023] Based on the solved positions and directions of all semantic elements in the spliced ​​semantic map, the spliced ​​semantic map is corrected.

[0024] Optionally, before the step of using the position and heading changes of the elements associated in the first semantic map as constraints and the step of using the position and heading changes of the elements associated in the second semantic map as constraints, the step further includes:

[0025] Obtaining the associated semantic elements in the first semantic map, and obtaining the associated semantic elements in the second semantic map;

[0026] The element position and heading change of the associated semantic element in the first semantic map is obtained based on the first global positioning data, and the element position and heading change of the associated semantic element in the second semantic map is obtained based on the second global positioning data.

[0027] Optionally, the step of solving the positions and directions of all semantic elements in the spliced ​​semantic map based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method includes:

[0028] Solving the position error and / or height error between corresponding semantic elements in the spliced ​​semantic map based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method;

[0029] Based on the solved pose error and / or height error, the positions and directions of all semantic elements in the spliced ​​semantic map are solved.

[0030] The present invention further provides a map generating device, comprising:

[0031] an acquisition module, configured to, upon detecting that a vehicle enters a parking lot and is powered off, acquire the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, wherein the first global positioning data is temporally associated with the first semantic map;

[0032] A construction module, configured to construct a second semantic map based on the position and heading information of the vehicle at the time of power-off when detecting that the vehicle starts to leave the parking lot;

[0033] a collection and association module, configured to collect second global positioning data after detecting that the global positioning signal is restored, and temporally associate the second global positioning data with the second semantic map;

[0034] The splicing module is configured to splice the first semantic map and the second semantic map based on the first global positioning data and the second global positioning data to obtain a spliced ​​semantic map.

[0035] Optionally, the device further comprises:

[0036] A correction module is used to correct the spliced ​​semantic map based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method.

[0037] The present invention also proposes a terminal device, which includes a memory, a processor, and a map generation program stored in the memory and executable on the processor. When the map generation program is executed by the processor, the steps of the map generation method described above are implemented.

[0038] The present invention further provides a computer-readable storage medium having a map generation program stored thereon. When the map generation program is executed by a processor, the steps of the map generation method described above are implemented.

[0039] The map generation method, apparatus, terminal device, and storage medium proposed in an embodiment of the present invention obtain, when detecting that a vehicle enters a parking lot and is powered off, the position and heading information of the vehicle at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, wherein the first global positioning data is temporally associated with the first semantic map, and the global positioning signal in the parking lot is lost; when detecting that the vehicle starts to leave the parking lot, a second semantic map is constructed based on the position and heading information of the vehicle at the time of power-off; after detecting that the global positioning signal is restored, the second global positioning data is collected, and the second global positioning data is temporally associated with the second semantic map; based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map. Therefore, by constructing maps for entering and exiting the parking lot, and collecting positioning data before and after the global positioning signal is lost and restored, map splicing is performed based on the first global positioning data and the second global positioning data, thereby reducing map construction errors and improving the accuracy of parking lot map construction; further, based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected, thereby combining map splicing, constraint rules and map fusion optimization method to correct the spliced ​​map, further reducing map construction errors and improving the accuracy of parking lot map construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a schematic diagram of the functional modules of the terminal device to which the map generating device of the present invention belongs;

[0041] Figure 2 This is a flowchart of a first embodiment of a map generation method according to the present invention;

[0042] Figure 3a This is a schematic diagram of the map construction effect when a vehicle enters a parking lot in an embodiment of the map generation method of the present invention;

[0043] Figure 3b This is a schematic diagram of the map construction effect after the vehicle leaves the parking lot in the map generation method embodiment of the present invention;

[0044] Figure 3c This is a schematic diagram of the effect of map splicing in an embodiment of the map generation method of the present invention;

[0045] Figure 4 This is a flow chart of a second embodiment of a map generation method according to the present invention;

[0046] Figure 5a This is an example diagram of the first semantic map in the map generation method embodiment of the present invention;

[0047] Figure 5b This is an example diagram of the second semantic map in the map generation method embodiment of the present invention;

[0048] Figure 6a A schematic top view of a semantic element provided by an embodiment of the present invention;

[0049] Figure 6b A schematic side view of a semantic element provided by an embodiment of the present invention;

[0050] Figure 7 This is a flowchart of a third embodiment of a map generation method according to the present invention;

[0051] Figure 8 This is a flow chart of a fourth embodiment of a map generation method according to the present invention;

[0052] Figure 9 Schematic diagram of the functional modules of an embodiment of a map generating device according to the present invention.

[0053] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0055] The main solution of an embodiment of the present invention is: when a vehicle is detected to enter a parking lot and power is turned off, the position and heading information of the vehicle at the time of power-off, as well as the first semantic map constructed before the vehicle is powered off and the first global positioning data collected are obtained, the first global positioning data and the first semantic map are temporally associated, and the global positioning signal in the parking lot is lost; when it is detected that the vehicle starts to leave the parking lot, a second semantic map is constructed based on the position and heading information of the vehicle at the time of power-off; after detecting that the global positioning signal is restored, the second global positioning data is collected, and the second global positioning data is temporally associated with the second semantic map; based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map. Therefore, by constructing maps for entering and exiting the parking lot, and collecting positioning data before and after the global positioning signal is lost and restored, map splicing is performed based on the first global positioning data and the second global positioning data, thereby reducing map construction errors and improving the accuracy of parking lot map construction; further, based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected, thereby combining map splicing, constraint rules and map fusion optimization method to correct the spliced ​​map, further reducing map construction errors and improving the accuracy of parking lot map construction.

[0056] Technical terms involved in the embodiments of the present invention:

[0057] AVP, short for Automated Valet Parking, is an application of autonomous driving in parking scenarios. It provides fully automated valet parking, ultimately aiming to replace traditional manual valet parking, saving users significant parking time and resolving the challenge of long parking queues during peak hours.

[0058] The on-site solution for finding parking spaces using the AVP function involves path planning, space identification, obstacle avoidance, and character recognition. Parking space identification requires the vehicle to autonomously identify various parking space types. Parking space types are typically categorized based on various factors, including direction, markings, and character recognition.

[0059] Obstacles to avoid mainly include static obstacles and dynamic obstacles. Common static obstacle types include: pillars, limit posts, wheel chocks, signs, ground locks, walls, trees, green belts, curbs, etc. Common dynamic obstacle types include: pedestrians, animals, non-motorized vehicles, motor vehicles, etc.

[0060] From the basic principles of AVP, we can see that the key technologies involved in AVP functions include high-precision maps, SLAM, fusion perception, fusion positioning, path planning, etc. Among them:

[0061] High-precision maps, also known as high-resolution maps (HD Maps), are maps specifically designed for autonomous driving. For the AVP function, HD maps primarily refer to parking lot HD maps: By formatting and storing various traffic elements within the parking lot, HD maps can provide vehicles with complete, highly accurate, and detailed information about the parking lot's interior, including high-precision coordinates, accurate parking spaces, aisles, pillars, signage, ground lines, and other information. The absolute accuracy of HD maps is generally within sub-meters. Taking Amap as an example, the absolute accuracy can reach within 10 centimeters, and the relative lateral accuracy is often even higher. HD maps store all traffic information in the parking lot and serve as the basis for vehicle route planning and positioning.

[0062] SLAM, short for Simultaneous Localization And Mapping, is a method for building high-precision maps. Its principle is that a vehicle, in a parking lot it has just arrived at, locates itself while moving based on its position and environmental perception. Simultaneously, it constructs an incremental map based on its positioning, enabling autonomous positioning and navigation.

[0063] Current mainstream SLAM relies primarily on visual semantics, using cameras to identify the surrounding environment, perform semantic analysis, confirm the current environment, and complete positioning and mapping. However, with the mass production of LiDAR in vehicles, laser SLAM may become the new mainstream.

[0064] Fusion perception and environmental awareness are paramount in autonomous driving, and are also key challenges for AVP. As we all know, different types of perception sensors, such as cameras, millimeter-wave radars, and lidars, each have their own strengths and weaknesses. Therefore, fusing the perception results of multiple sensors to leverage their complementary strengths can improve perception precision and accuracy. Currently, multi-sensor fusion has become a trend in autonomous driving. By fusing the perception results of multiple cameras and radars, the vehicle can more accurately identify its surroundings and reconstruct scenes. For AVP, perception results from sensors such as front-view cameras, side-view cameras, surround-view cameras, millimeter-wave radars, lidars, and ultrasonic radars can all be fused. The fused perception results significantly enhance environmental recognition.

[0065] Fusion positioning: The positioning of traditional vehicles mainly relies on GNSS, the global navigation satellite system, such as GPS and Beidou.

[0066] Path planning refers to planning the parking route for a vehicle as it enters a parking space. Parking path planning is a relatively complex problem, involving obstacle and vehicle trajectory prediction, drivable area selection, local trajectory planning, and vehicle control. Different planning algorithms can produce completely different automated parking results.

[0067] This embodiment takes into account that the AVP function requires a parking map. Although current smart vehicles have sufficient positioning sensors and computing units to construct parking maps, most parking maps are located indoors, such as in underground garages. Underground garages lack satellite signal coverage, and traditional GNSS (Global Navigation Satellite System) positioning fails. When GNSS signals are lost, the position estimation error during map construction increases, resulting in inaccurate absolute positions on the map.

[0068] Therefore, the present invention proposes a solution. By constructing maps for entering and exiting the parking lot, and collecting positioning data before and after the global positioning signal is lost and restored, and combining map splicing, constraint rules and map fusion optimization methods, the spliced ​​map is corrected. This can achieve accurate positioning of the vehicle's position in the parking lot, reduce map construction errors, and improve the accuracy of parking lot map construction.

[0069] Specifically, refer to Figure 1 , Figure 1 This is a schematic diagram of the functional modules of a terminal device to which the map generation device of the present invention belongs. The map generation device can be a device independent of the terminal device, capable of performing operations such as visual perception and data processing, and can be hosted on the terminal device in the form of hardware or software. The terminal device can be a smart terminal with data processing capabilities, such as a mobile phone, tablet computer, or in-vehicle device. It can also be a terminal device or server with data processing capabilities, such as a vehicle.

[0070] In this embodiment, the terminal device to which the map generating apparatus belongs includes at least an output module 110 , a processor 120 , a memory 130 and a communication module 140 .

[0071] The memory 130 stores an operating system and a map generation program. The output module 110 may be a display screen, etc. The communication module 140 may include a WIFI module, a mobile communication module, and a Bluetooth module, etc., and communicates with an external device or server through the communication module 140.

[0072] When the map generation program in the memory 130 is executed by the processor, the following steps are implemented:

[0073] Upon detecting that a vehicle enters a parking lot and is powered off, obtaining the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, wherein the first global positioning data is temporally associated with the first semantic map, and a global positioning signal is lost within the parking lot;

[0074] When detecting that a vehicle starts to leave the parking lot, constructing a second semantic map based on the vehicle's position and heading information at the time of power-off;

[0075] After detecting that the global positioning signal has been restored, collecting second global positioning data, and temporally associating the second global positioning data with the second semantic map;

[0076] Based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map.

[0077] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0078] Based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected.

[0079] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0080] In a startup state before the vehicle enters the parking lot, continuously constructing a first semantic map of a preset area and collecting first global positioning data;

[0081] Associating the first global positioning data with corresponding track points on the first semantic map in time;

[0082] After detecting that the global positioning signal is lost, the first semantic map continues to be constructed until it is detected that the vehicle enters the parking lot and is powered off.

[0083] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0084] Determining whether the accuracy of the second global positioning data meets a preset standard;

[0085] If the accuracy of the second global positioning data reaches a preset standard, the step of splicing the first semantic map and the second semantic map is performed.

[0086] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0087] Obtaining an overlapping area between the first semantic map and the second semantic map in the spliced ​​semantic map, and using matching semantic elements in the overlapping area as mutual constraints; and / or using position and heading changes of associated elements in the first semantic map as constraints, and using position and heading changes of associated elements in the second semantic map as constraints;

[0088] Based on the first global positioning data and the second global positioning data, acquiring corresponding constrained positioning data;

[0089] Based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method, the positions and directions of all semantic elements in the spliced ​​semantic map are obtained;

[0090] Based on the solved positions and directions of all semantic elements in the spliced ​​semantic map, the spliced ​​semantic map is corrected.

[0091] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0092] Obtaining the associated semantic elements in the first semantic map, and obtaining the associated semantic elements in the second semantic map;

[0093] The element position and heading change of the associated semantic element in the first semantic map is obtained based on the first global positioning data, and the element position and heading change of the associated semantic element in the second semantic map is obtained based on the second global positioning data.

[0094] Furthermore, when the map generation program in the memory 130 is executed by the processor, the following steps are also implemented:

[0095] Solving the position error and / or height error between corresponding semantic elements in the spliced ​​semantic map based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method;

[0096] Based on the solved pose error and / or height error, the positions and directions of all semantic elements in the spliced ​​semantic map are solved.

[0097] Through the above solution, this embodiment, when detecting that a vehicle has entered a parking lot and powered off, obtains the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle powered off. The first global positioning data is temporally associated with the first semantic map, and the global positioning signal in the parking lot is lost. When detecting that the vehicle has started to leave the parking lot, a second semantic map is constructed based on the vehicle's position and heading information at the time of power-off. After detecting that the global positioning signal has been restored, second global positioning data is collected and temporally associated with the second semantic map. The first and second semantic maps are spliced ​​together based on the first and second global positioning data to obtain a spliced ​​semantic map, thereby reducing map construction errors and improving the accuracy of parking lot map construction. Furthermore, the spliced ​​semantic map is corrected based on the first and second global positioning data, as well as preset constraint rules, in combination with a preset map fusion optimization method. Thus, by combining map splicing, constraint rules, and map fusion optimization method, the spliced ​​map is corrected, further reducing map construction errors and improving the accuracy of parking lot map construction.

[0098] Based on the above terminal device architecture but not limited to the above architecture, an embodiment of the method of the present invention is proposed.

[0099] The execution subject of the method of this embodiment can be a map generating device, or a vehicle-mounted device, server or other terminal device with a map generating function. This embodiment takes a map generating device as an example, and the map generating device can be integrated into a vehicle.

[0100] The solution of this embodiment can be applied to parking lots or similar places without GNSS environments. This embodiment takes a parking lot as an example.

[0101] Reference Figure 2 , Figure 2 This is a flow chart of a first embodiment of a map generation method according to the present invention. The map generation method includes:

[0102] Step S101: When a vehicle is detected entering a parking lot and powered off, obtaining the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed before the vehicle was powered off and first global positioning data collected, wherein the first global positioning data is temporally associated with the first semantic map, and the global positioning signal is lost within the parking lot;

[0103] Among them, vehicle power-off may refer to the vehicle being in an off state after entering a parking lot, for example, by turning off the vehicle engine after finding a parking space and completing parking, so that the vehicle is in a power-off state. The vehicle power-off state may correspond to the vehicle startup state.

[0104] The parking lot referred to in this embodiment is typically an underground garage. As previously mentioned, satellite signal coverage is lost in underground garages, rendering traditional GNSS (Global Navigation Satellite System) positioning ineffective. This embodiment is implemented in scenarios where global positioning signals are lost within a parking lot. These global positioning signals may include GNSS signals, among others.

[0105] Taking GNSS signals as an example, when GNSS signals are lost, position estimation errors during map construction will become increasingly larger, resulting in inaccurate absolute positions on the map. This embodiment constructs maps for vehicles entering and exiting parking lots, simultaneously collecting positioning data before and after global positioning signals are lost. Combined with map stitching, constraint rules, and map fusion optimization methods, the stitched map is corrected, thereby reducing map construction errors and improving the accuracy of parking lot map construction.

[0106] Specifically, in this embodiment, before the vehicle is powered off, a first semantic map is pre-constructed and first global positioning data is collected at the same time. The first global positioning data is temporally associated with the first semantic map.

[0107] Among them, the first semantic map can be realized through composition technologies such as SLAM and high-precision maps. For example, the surrounding environment can be identified through perception sensors such as cameras and radars on the vehicle. After semantic analysis, the current environment can be confirmed, and positioning and mapping can be completed. Specifically, visual perception technology can be combined to detect the position and direction of semantic elements such as parking spaces, pillars, walls, arrows, roadblocks, speed bumps, lane lines, etc. in the parking lot, and the map trajectory can be obtained through dead reckoning through inertial navigation equipment and wheel speed sensors to construct the map.

[0108] The collection of the first global positioning data can be achieved through GNSS positioning or map positioning technology, and the map positioning technology can include SLAM positioning, high-precision map positioning, etc.

[0109] Among them, as an implementation method, in the starting state before the vehicle enters the parking lot, a first semantic map of a preset area is continuously constructed, and first global positioning data is collected at the same time; the first global positioning data is temporally associated with the corresponding trajectory points on the first semantic map.

[0110] Among them, the preset area range can be selected and set according to actual conditions, and the accuracy requirements of the positioning data can also be considered. For example, it can be set in a preset area near the parking lot, or in an area within a preset distance from the parking lot.

[0111] Among them, when collecting the first global positioning data, it is necessary to temporally associate the first global positioning data with the trajectory points of the second semantic map, that is, the semantic elements corresponding to the trajectory points in the constructed map are associated with the global positioning data collected at the current moment to ensure that the positioning data and the information of the map semantic elements are synchronized.

[0112] After detecting that the global positioning signal is lost, the first semantic map continues to be constructed until it is detected that the vehicle enters the parking lot and is powered off.

[0113] When a vehicle is detected entering a parking lot and powered off, the vehicle's position and heading information at the time of power-off is obtained, as well as a first semantic map constructed before the vehicle is powered off and first global positioning data collected, where the first global positioning data is temporally associated with the first semantic map.

[0114] The vehicle's position and heading information at the time of power-off refers to the position coordinates and direction information of the trajectory point at which the vehicle is located at the time of power-off, which can be obtained by dead reckoning using an inertial navigation device and a wheel speed sensor.

[0115] like Figure 3a As shown, when the vehicle is parked and powered off in the parking lot, the current position and heading of the vehicle are saved, and the first semantic map constructed by the vehicle is also saved ( Figure 3a The black track in the middle is the movement track of the vehicle when entering the parking lot) and the associated GNSS positioning or map positioning data (such as Figure 3a (the track point on the left side of the center entering the parking lot).

[0116] Step S102, when detecting that a vehicle starts to leave the parking lot, constructing a second semantic map based on the vehicle's position and heading information at the time of power-off;

[0117] When it is detected that the vehicle starts to leave the parking space, a second semantic map is constructed based on the position and heading information of the vehicle at the time of power-off until the vehicle leaves the parking lot.

[0118] The second semantic map is constructed in the same way as the first. This can be achieved through composition technologies such as SLAM and high-precision maps. For example, the surrounding environment can be identified through perception sensors such as cameras and radars on the vehicle. After semantic analysis, the current environment is confirmed and positioning and mapping are completed. Specifically, visual perception technology can be combined to detect the position and direction of semantic elements such as parking spaces, pillars, walls, arrows, and lane markings in the parking lot. The map trajectory is calculated through dead reckoning using inertial navigation equipment and wheel speed sensors to construct the map.

[0119] The second semantic map constructed can be Figure 3b As shown, Figure 3b The white track in the middle is the moving track of the vehicle.

[0120] Step S103: After detecting that the global positioning signal is restored, collecting second global positioning data, and temporally associating the second global positioning data with the second semantic map;

[0121] When the vehicle leaves the parking space and prepares to leave the parking lot, the vehicle's global positioning signal is detected in real time.

[0122] Typically, in underground parking garages, where there is no satellite signal coverage, the global positioning signal is lost. The global positioning signal is restored when the vehicle reaches or exits the parking garage exit.

[0123] After detecting that the global positioning signal is restored, second global positioning data is collected, and the second global positioning data is temporally associated with the second semantic map.

[0124] The second global positioning data is collected in the same manner as the first global positioning data, and can be achieved through GNSS positioning or map positioning technology, such as SLAM positioning and high-precision map positioning.

[0125] Taking GNSS signals as an example, when a vehicle enters a parking lot and before the GNSS signal is lost, the first global positioning data can be collected through GNSS positioning technology, such as Figure 3a As shown in the trajectory point of the parking lot; in the process of the vehicle leaving the parking lot, after the GNSS signal is restored, the second global positioning data can be collected by GNSS positioning technology, such as Figure 3b Track points at the exit from the parking lot are shown.

[0126] Among them, when collecting the second global positioning data, it is necessary to temporally associate the second global positioning data with the trajectory points of the second semantic map, that is, the semantic elements corresponding to the trajectory points in the constructed map are associated with the global positioning data collected at the current moment to ensure that the positioning data and the information of the map semantic elements are synchronized.

[0127] In addition, the collection area range of the second global positioning data can be selected and set according to actual conditions, and the accuracy requirements of the positioning data can also be considered. For example, it can be set in a preset area near the parking lot, or in an area within a preset distance from the parking lot.

[0128] Step S104: splicing the first semantic map and the second semantic map based on the first global positioning data and the second global positioning data to obtain a spliced ​​semantic map;

[0129] Among them, as an implementation method, the first semantic map and the second semantic map can be rigidly spliced. When splicing, the semantic elements in the map can be referred to, and the semantic elements can be used as a reference standard. Combined with the collected first global positioning data and second global positioning data, splicing is performed based on information such as the position and direction of the semantic elements to obtain a spliced ​​semantic map.

[0130] Through the above-mentioned solution, this embodiment, when detecting that a vehicle enters a parking lot and powers off, obtains the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle powers off. The first global positioning data is temporally associated with the first semantic map, and the global positioning signal in the parking lot is lost. When detecting that the vehicle starts to leave the parking lot, a second semantic map is constructed based on the vehicle's position and heading information at the time of power-off. After detecting that the global positioning signal has been restored, the second global positioning data is collected and temporally associated with the second semantic map. The first semantic map and the second semantic map are spliced ​​together based on the first and second global positioning data to obtain a spliced ​​semantic map. Thus, by constructing maps for entering and exiting the parking lot, while collecting positioning data before and after the global positioning signal is lost, and performing map splicing based on the first and second global positioning data, map construction errors are reduced and the accuracy of parking lot map construction is improved.

[0131] Reference Figure 4 , Figure 4 This is a flow chart of the second embodiment of the map generation method of the present invention. Figure 2 Based on the embodiment shown, the map generation method of this embodiment further includes:

[0132] Step S105 : Based on the first global positioning data, the second global positioning data, and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected.

[0133] Compared with the above Figure 2 The embodiment shown in FIG. 1 also includes a solution for correcting the spliced ​​semantic map.

[0134] Constraints can be defined by using GNSS positioning or outdoor map positioning data to constrain corresponding map track points. The concept of constraints is to ensure that the position coordinates and directions of the track points of each matching or associated semantic element on the map are absolutely accurate. Constraints can be set from the following three aspects:

[0135] On the one hand, if the first semantic map and the second semantic map have overlapping areas or common areas, the matching semantic elements in the overlapping areas or common areas are used as mutual constraints;

[0136] On the other hand, the position and heading changes of the associated semantic elements in the first semantic map are used as constraints;

[0137] On the other hand, the position and heading changes of the associated semantic elements in the second semantic map are used as constraints.

[0138] Therefore, based on the first global positioning data, the second global positioning data and the preset constraint rules, the positions and directions of all semantic elements in the spliced ​​semantic map are solved by a map fusion optimization method, and the spliced ​​semantic map is corrected.

[0139] Among them, the data to be processed by the map fusion optimization method is the position and direction of the trajectory points of all semantic elements on the spliced ​​semantic map. The principle of the optimization method is to change the coordinate data of the trajectory points on the entire map by solving the minimum error value to accurately locate the position coordinates and direction of each trajectory point on the map.

[0140] After the calculation is completed, save the processed map to obtain a map with higher absolute position accuracy. The processed semantic map can be referenced Figure 3c As shown, compared Figure 3b and Figure 3c It can be seen that Figure 3b The trajectory points of the second global positioning data at the end of the white trajectory are distributed on one side of the white trajectory. After correction, the trajectory points of the second global positioning data are distributed on both sides of the white trajectory, such as Figure 3c As shown, it can be seen that the positioning accuracy of the trajectory points of the map semantic elements is improved by correcting the map fusion optimization method.

[0141] Specifically, as an implementation method, the step of correcting the spliced ​​semantic map based on the first global positioning data, the second global positioning data, and preset constraint rules, and in combination with a preset map fusion optimization method, may include:

[0142] First, obtaining an overlapping area between the first semantic map and the second semantic map in the spliced ​​semantic map, and using the semantic elements that match each other in the overlapping area as mutual constraints; and / or using the position and heading changes of the elements associated in the first semantic map as constraints, and using the position and heading changes of the elements associated in the second semantic map as constraints;

[0143] Among them, before the step of using the associated element position and heading changes in the first semantic map as constraints and the step of using the associated element position and heading changes in the second semantic map as constraints, the associated semantic elements in the first semantic map and the associated semantic elements in the second semantic map can be obtained; the element position and heading changes of the associated semantic elements in the first semantic map are obtained based on the first global positioning data, and the element position and heading changes of the associated semantic elements in the second semantic map are obtained based on the second global positioning data.

[0144] The element position and heading change refers to the difference between the position and heading of two semantic elements, such as the difference in position coordinates or the difference in direction angles.

[0145] Then, based on the first global positioning data and the second global positioning data, corresponding constrained positioning data is acquired;

[0146] Then, based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method, the positions and directions of all semantic elements in the spliced ​​semantic map are solved;

[0147] The map fusion optimization method may adopt a conventional optimization method, which is not specifically limited in this embodiment, as long as the positioning accuracy of the map data is achieved.

[0148] Among them, as an implementation method, in a scheme for solving the positions and directions of all semantic elements in the spliced ​​semantic map based on the positioning data of the corresponding constraints and in combination with a preset map fusion optimization method, the posture error and / or height error between the corresponding semantic elements in the spliced ​​semantic map can be solved based on the positioning data of the corresponding constraints and in combination with a preset map fusion optimization method, and the positions and directions of all semantic elements in the spliced ​​semantic map can be solved based on the posture error and / or height error obtained by the solution.

[0149] Finally, based on the positions and directions of all semantic elements in the spliced ​​semantic map obtained by solving the spliced ​​semantic map, the spliced ​​semantic map is corrected.

[0150] This embodiment uses the above solution to construct maps for parking lot entrances and exits, while also collecting positioning data before and after global positioning signal loss. The spliced ​​map is then corrected by combining map splicing, constraint rules, and map fusion optimization methods, further reducing map construction errors and improving the accuracy of parking lot map construction.

[0151] The following describes in detail the principle of correcting and optimizing the spliced ​​semantic map in the map generation solution of this embodiment through the implementation process of a specific map fusion optimization method.

[0152] After obtaining the first semantic map and the second semantic map, at least two first semantic elements in the first semantic map may be determined, for example, Figure 5a As shown, the parking space information a and lane line information b in the first semantic map can be determined, such as Figure 5b As shown, the parking space information A and lane line information B in the second semantic map are determined, wherein the information of the first semantic map is indicated by lowercase letters, and the information of the second semantic map is indicated by uppercase letters.

[0153] After determining the two semantic elements of the first semantic map, a second semantic element matching the first semantic element can be determined in the second semantic map. For example, based on the parking space information a and lane line information b in the first semantic map, parking space information A matching the parking space information a and lane line information B matching the lane line information b can be determined from multiple semantic elements in the second semantic map.

[0154] Thus, the semantic elements that match each other in the overlapping area or common area of ​​the first semantic map and the second semantic map can be used as mutual constraints.

[0155] Then, determining first relative information between at least two first semantic elements, and determining second relative information between at least two second semantic elements;

[0156] The relative information may be two-dimensional relative information or three-dimensional relative information between two semantic elements.

[0157] Specifically, after determining the matching semantic element, the two-dimensional and three-dimensional information of the semantic element can be determined. For example, a three-dimensional coordinate system in any semantic map can be determined, and then the coordinate information of the semantic element in the semantic map can be determined based on the coordinate information. The two-dimensional and three-dimensional information of the semantic element can be determined based on the collected first and second global positioning data.

[0158] After determining the two-dimensional information and three-dimensional information of the semantic element, the two-dimensional relative information and three-dimensional relative information of the two semantic elements can be determined based on the two-dimensional information and three-dimensional information of the semantic element to obtain the first relative information of the first semantic element and the second relative information of the second semantic element.

[0159] For example, the two-dimensional information and three-dimensional information of parking space information a and the two-dimensional information and three-dimensional information of lane line information can be determined, and then the two-dimensional relative information and three-dimensional relative information between parking space information a and lane line information b, that is, the first relative information, can be determined.

[0160] Thus, the position and heading changes of the associated semantic elements in the first semantic map can be used as constraints, and the position and heading changes of the associated semantic elements in the second semantic map can be used as constraints. The position and heading changes of the semantic elements can be determined based on the two-dimensional relative information and the three-dimensional relative information of the semantic elements.

[0161] Then, optimizing the first semantic map and the second semantic map according to the first relative information and the second relative information;

[0162] After obtaining the first relative information and the second relative information, the relative relationship between the first semantic elements, the relative relationship between the second semantic elements, and the error between the first semantic element and the second semantic element can be determined based on the first relative information and the second relative information, such as the error between the two-dimensional information of parking space information a and parking space information A, the two-dimensional relative information between parking space information a and lane line information b, and the error between the two-dimensional relative information between parking space information A and lane line information B.

[0163] After the relative relationship and the error between the first semantic element and the second semantic element are determined, the two-dimensional information and the three-dimensional information of the semantic elements may be optimized and adjusted according to the relative relationship and the error.

[0164] Finally, the optimized first semantic map and the optimized second semantic map are fused.

[0165] Therefore, based on the first global positioning data, the second global positioning data and the preset constraint rules, the positions and directions of all semantic elements in the spliced ​​semantic map are solved by the map fusion optimization method, and the spliced ​​semantic map is fused and corrected.

[0166] After optimizing the semantic map, different fusion times can be determined, and then the two semantic maps can be weighted and fused according to the different fusion times. By determining the number of map fusion times and determining the difficulty of map fusion based on the fusion times, and then weighting and fusing the semantic maps according to the difficulty, the accuracy of map fusion can be improved, making the fused map more accurate.

[0167] In this embodiment, the implementation principle of the map fusion optimization method is as follows:

[0168] The constructed first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map.

[0169] Based on the spliced ​​semantic map, matching semantic elements in the overlapping area or common area of ​​the first semantic map and the second semantic map are determined. After determining the matching semantic elements, the two-dimensional information of the semantic elements, such as posture information, and the three-dimensional information of the semantic elements, such as height information, can be determined.

[0170] For example, based on the collected global positioning data, the three-dimensional coordinate system in the semantic map can be determined, and then the coordinate information of the semantic elements in the semantic map can be determined according to the coordinate system, and the posture information and height information of the semantic elements can be determined according to the coordinate information.

[0171] In practical applications, the rotation information of the semantic element relative to the x-axis and the y-axis may also be determined, and the rotation of the semantic element may be optimized based on the rotation information relative to the x-axis and the y-axis.

[0172] However, since there is no large error between the rotation information of the semantic element and the x-axis and the y-axis, optimizing the rotation information will not bring a higher improvement in the quality of map fusion. Moreover, the optimization of the rotation information requires a large amount of calculation, a complex processing method, and high hardware requirements. Therefore, only the height information of the semantic element can be optimized, which simplifies the optimization process, reduces the amount of calculation, and improves the accuracy of the height optimization by determining the relative height information, height error value and other information.

[0173] After determining the position information and height information of the semantic elements, the relative position information between the two semantic elements can be determined based on the position information of the semantic elements, and the relative height information between the two semantic elements can be determined based on the height information of the semantic elements.

[0174] like Figure 6a As shown, the pose information of semantic elements 1 to n may include position information and angle information (equivalent to direction information). The position information and angle information of the semantic elements can be determined based on the coordinate information, and then converted into the pose information of the semantic elements based on the position information and angle information. Specifically, the following matrix conversion can be used:

[0175]

[0176]

[0177] Among them, matrix (1) can be expressed as a value for converting the position and posture of semantic element i in three-dimensional map data, and matrix (2) can be expressed as a value for converting the position and posture of semantic element j in three-dimensional map data. Semantic element i and semantic element j can be semantic elements matched in the first semantic map and the second semantic map, or any two semantic elements in the same semantic map. For matrix (1), T i It can be expressed as the absolute position of semantic element i, Can be the angle information of semantic element i, tx i can be the x-axis coordinate information of semantic element i in the coordinate system, ty i It can be the y-axis coordinate information of the semantic element i in the coordinate system, and the same applies to matrix (2).

[0178] In one embodiment of the present invention, the height information of semantic elements 1 to n can be as follows: Figure 6b As shown, the height information of the semantic element can be determined according to the coordinate information, wherein z1 can be represented as the height information of the semantic element 1.

[0179] As mentioned above, after determining the position information and height information of the semantic elements, the relative position information between two semantic elements can be determined based on the position information of the semantic elements, and the relative height information between two semantic elements can be determined based on the height information of the semantic elements.

[0180] Thus, the first relative position information and the first relative height information of the first semantic element and the second relative position information and the second relative height information of the second semantic element are obtained.

[0181] After obtaining the relative posture information and relative height information, the relative relationship between the first semantic elements, the relative relationship between the second semantic elements, and the error between the first semantic element and the second semantic element can be determined based on the relative posture information and relative height information, such as the error between the posture information of parking space information a and parking space information A, the relative posture information between parking space information a and lane line information b, and the error between the relative posture information of parking space information A and lane line information B.

[0182] The error between the first semantic element and the second semantic element may be a relative posture error between first relative posture information of the first semantic element and second relative posture information of the second semantic element.

[0183] After determining the relative relationship and the error between the first semantic element and the second semantic element, the target posture information and target height information of the first semantic element can be determined based on the relative relationship and the error, as well as the target posture information and target height information of the second semantic element can be determined, so that the posture information and height information of the semantic elements can be optimized and adjusted respectively according to the target posture information and target height information.

[0184] The following details the determination of the relative pose information between two semantic elements, as well as the calculation process of the relative pose error and relative height error:

[0185] After determining the pose information of the semantic elements, the relative pose information between the two semantic elements can be determined based on the pose information of the semantic elements. Specifically, the relative pose information can be determined using the following matrix:

[0186]

[0187] Among them, subscript i and subscript j can be represented as semantic element i and semantic element j, T ij It can be the relative position of semantic element j relative to semantic element i, T i -1 Can be the inverse matrix of semantic element i, R j Can be the transformation matrix of the angle information of semantic element j, It can be the transposed matrix of the transformation matrix of the angle information of semantic element i, t j It can be the position of semantic element j.

[0188] In one embodiment of the present invention, when semantic element i and semantic element j are any two semantic elements in the same semantic map, the relative position information of the two semantic elements can be T ij Before the optimization, the relative position between the two semantic elements can be seen through the wavy line and T ij Distinguish, specifically can be expressed as the following matrix:

[0189]

[0190] in, It can be used to provide the relative angle information of two semantic elements before optimization. It can be used to transform the relative angle information of two semantic elements before optimization. The relative position of two semantic elements can be optimized before they are found.

[0191] In one embodiment of the present invention, when semantic element i and semantic element j are matched semantic elements in two semantic map data, the relative position information of the two semantic elements can be T ij ,and It can be the identity matrix, which can be expressed as the following matrix:

[0192]

[0193] in, Can be T ij The identity matrix, The relative angle information of the two semantic elements before optimization can be The conversion matrix of the relative angle information of the two semantic elements before optimization may be used.

[0194] After determining the relative pose information, the pose error value can be determined based on the relative pose information. Specifically, the pose error value can be determined by the following matrix:

[0195]

[0196] in, Can be T ij The inverse matrix of It can be the transposed matrix of the transformation matrix of the relative angle information of two semantic elements, It can be the transposed matrix of the conversion matrix of the angle information of the semantic element i.

[0197] After determining the matrix (5), the matrix (5) can be converted into the following matrix:

[0198]

[0199] Among them, e ij It can be expressed as the posture error value of the first relative posture information and the second relative posture information. In one embodiment of the present invention, when the semantic element i and the semantic element j are any two semantic elements in the same semantic map, e ij It can also be expressed as a posture error value of the first relative posture information or the second relative posture information.

[0200] Then, the pose information of the two first semantic elements and the two second semantic elements and their correlation with the relative pose error may be determined according to the pose error value;

[0201] Among them, the relative pose error and pose information can be independent variables in the association relationship, the relative pose error can be the dependent variable in the association relationship, and the pose error value is a known value of the relative pose error. After determining the pose error value, the association relationship between the relative pose error and the pose information of the first semantic element and / or the second semantic element can be determined based on the pose error value and the corresponding first relative pose information and second relative pose information. The algorithm for this association relationship can be the map fusion optimization method described in this embodiment, and the association relationship can be the following formula:

[0202]

[0203] Among them, f can be expressed as the correlation between relative pose error and pose information, e k_ij Can be the k-th pose error value, e k_ij It can be defined as a column vector, e k_ij T Can be the transposed matrix of the k-th pose error value, w kIt can be expressed as map maturity, which corresponds to the number of map fusions. Then, with the goal of minimizing the relative pose error, target pose information of the two first semantic elements and the two second semantic elements is determined.

[0204] After determining the correlation formula between the relative posture error and the posture information, since the correlation formula can be expressed as the correlation between the relative posture error and the posture information, the target posture information corresponding to the minimum relative posture error can be determined with the goal of minimizing the relative posture error.

[0205] In practical applications, the first relative posture information and / or the second relative posture information corresponding to the minimum value of the correlation formula, that is, when the value of f is the minimum, can be calculated, and then the corresponding position information and angle information can be determined. The position information and angle information can be determined by the following formula:

[0206]

[0207]

[0208] Among them, β i It can be expressed as a matrix of position information and angle information of semantic element i, and the subscripts 1 to n of β can represent the position information and angle information of semantic elements 1 to n.

[0209] Furthermore, the process of optimizing and adjusting the height information of the semantic element based on the target height information is as follows:

[0210] determining a height error value according to first relative height information of the first semantic element and second relative height information of the second semantic element;

[0211] The height error value may be an error value between the first relative height information and the second relative height information.

[0212] Specifically, first, after determining the height information of the semantic elements, any two semantic elements in one of the semantic maps can be determined as a semantic element group, and the relative height information of any group of semantic element groups can be determined based on the height information of the semantic elements. Similarly, the relative height information of any group of semantic element groups in another semantic map can also be determined, and then the relative height error value can be determined based on the relative height information in different semantic maps. The relative height error value can be determined by the following formula:

[0213]

[0214] Among them, e zij It can be expressed as a relative height error value, z j and z iIt can be expressed as the height information of semantic element j and semantic element i in one of the semantic map data. The relative height information of the semantic element group composed of semantic element i and semantic element j in the semantic map can be obtained by difference, that is, z j -z i , It can be expressed as the relative height information before optimization of the semantic element group consisting of the matching semantic element i and the matching semantic element j in another semantic map.

[0215] Then, determining, based on the relative height error value, an association relationship between height information of at least two first semantic elements and at least two second semantic elements and the relative height error value;

[0216] After determining the relative height error value, an association relationship between the relative height error and the height information of the first semantic element and the second semantic element can be determined based on the relative height error value and the corresponding first relative height information and the second relative height information. The association relationship can be expressed as follows:

[0217]

[0218] Among them, f z It can be expressed as the correlation between relative height error and height information, e zk_ij It can be expressed as the height error between the kth semantic element group in the first semantic element and the kth semantic element group in the second semantic element, e zk_ij T It can be expressed as the transposed matrix of the height error value of the kth semantic element group in the first semantic element and the kth semantic element group in the second semantic element. k can be used to mark the semantic element group in the first semantic element and the second semantic element. The value of k has no effect on the association relationship formula. m can be the total number of semantic element groups in the first semantic element or the second semantic element. w k It can be expressed as map maturity, which can correspond to the number of map fusions.

[0219] For example, z1_12 It can be expressed as the relative height error value between the first group of semantic elements in the first semantic element and the first group of semantic elements in the second semantic element. The first group of semantic elements can include semantic element 1 and semantic element 2, that is, semantic element 1 and semantic element 2 in the first semantic element, and semantic element 1 and semantic element 2 in the second semantic element.

[0220] Then, with the goal of minimizing the relative height error, target height information of at least two first semantic elements and at least two second semantic elements is determined.

[0221] After determining the correlation formula, since the correlation formula can be expressed as the correlation between the relative height error and the height information, the minimum relative height error can be taken as the goal, and the value of the correlation formula between the relative height error and the height information is calculated to be the minimum, that is, when f z When the value of is the smallest, the corresponding first relative height information and the second relative height information are obtained, and the corresponding target height information can be determined.

[0222] Among them, f can be obtained by determining the gradient of the association formula z For example, when the gradient is 0, f z The value of is the smallest, and the target height information can be determined. Specifically, it can be calculated using the following formula:

[0223] Kz=b

[0224] Among them, z can be the target height information, b can be The situation expressed, Can be a gradient, It can be the partial derivative for the height information in the gradient, It can be expressed as z = 0

[0225] As an example, the gradient can be represented by the following matrix:

[0226]

[0227] in, It can be the partial derivative of the height information of semantic element 1 in the relationship formula between relative height error and height information. The subscripts 1 to n of z can represent the height information of semantic elements 1 to n. k T Can be e zk_ij T , Can be targeted at e k The partial derivative of the height information of semantic element 1 in .

[0228] From the above gradient, we can see that Can be converted to Substituting Kz=b into the above equation, the z value can be directly solved, that is, the target height information can be calculated.

[0229] Finally, after obtaining the target information, such as the target posture information and the target height information, the posture information and the height information of the semantic element can be optimized and adjusted according to the target information.

[0230] Map fusion is performed on the optimized first semantic map and the optimized second semantic map.

[0231] In the above solution, when the two semantic maps are spliced ​​and fused, at least two semantic elements in the first semantic map may be determined, and it may be determined whether there is a semantic element in the second semantic map that matches the first semantic map.

[0232] After determining the matching semantic elements, the overall position information of the matching semantic elements can be determined in different semantic maps respectively, and then the transformation matrix for different semantic maps can be generated according to the different overall position information. After determining the transformation matrix, the corresponding transformation matrix can be used to make overall adjustments to the different semantic maps.

[0233] After obtaining the adjusted semantic map, the relative pose information between the two semantic elements can be determined based on the position information and angle information of the semantic elements, and the value of the relative pose error, that is, the pose error value, can be determined based on the relative pose information. Then, the correlation between the pose information and the relative pose error can be determined based on the relative pose information and the pose error value, and the target pose information can be determined with the minimum relative pose error as the goal to perform planar pose optimization.

[0234] After plane pose optimization, the relative height information between two semantic elements can be determined based on the height information of the semantic elements, and the relative height error value can be determined based on the relative height information. Then, the correlation between the height information and the relative height error can be determined based on the relative height information and the relative height error value, and the target height information can be determined with the minimum relative height error as the goal to perform height optimization.

[0235] After optimizing the semantic map, the fusion times of different semantic maps can be determined, and then the two semantic maps can be weightedly fused according to the different fusion times.

[0236] Finally, the unmatched semantic elements in the two semantic maps are fused, that is, the semantic elements of one semantic map are added to the other semantic map.

[0237] Through the above scheme, this embodiment obtains the position and heading information of the vehicle at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, when it is detected that the vehicle enters the parking lot and is powered off. The first global positioning data is temporally associated with the first semantic map, and the global positioning signal in the parking lot is lost. When it is detected that the vehicle starts to leave the parking lot, a second semantic map is constructed based on the position and heading information of the vehicle at the time of power-off. After detecting that the global positioning signal is restored, the second global positioning data is collected and temporally associated with the second semantic map. The first semantic map and the second semantic map are spliced ​​together to obtain a spliced ​​semantic map. The spliced ​​semantic map is corrected based on the first global positioning data, the second global positioning data, and preset constraint rules, and in combination with a preset map fusion optimization method. Therefore, by constructing maps for entering and exiting the parking lot, collecting positioning data before and after the global positioning signal is lost and restored, and combining map stitching, constraint rules and map fusion optimization methods, the stitched map is corrected, thereby reducing map construction errors and improving the accuracy of parking lot map construction. In addition, the fusion scheme in the map generation scheme is simple, which improves the accuracy and efficiency of map fusion.

[0238] Reference Figure 7 , Figure 7 This is a flow chart of the third embodiment of the map generation method of the present invention. Figure 4 Based on the embodiment shown, in the above step S101, when it is detected that a vehicle enters a parking lot and is powered off, obtaining the vehicle's position and heading information at the time of power-off, as well as the first semantic map constructed before the vehicle is powered off and the first global positioning data collected, further includes:

[0239] Step S1001: in a startup state before the vehicle enters a parking lot, continuously constructing a first semantic map of a preset area and collecting first global positioning data;

[0240] Step S1002: temporally associating the first global positioning data with corresponding trajectory points on the first semantic map;

[0241] Step S1003: After detecting that the global positioning signal is lost, continue to construct the first semantic map until it is detected that the vehicle enters the parking lot and is powered off.

[0242] Compared with the above Figure 4 The illustrated embodiment also includes a solution for constructing a first semantic map of a preset area and collecting first global positioning data.

[0243] Specifically, before the vehicle is powered off, the first semantic map is continuously constructed, and the first global positioning data is collected at the same time, where the first global positioning data is temporally associated with the first semantic map.

[0244] Among them, the first semantic map can be realized through composition technologies such as SLAM and high-precision maps. For example, the surrounding environment can be identified through perception sensors such as cameras and radars on the vehicle. After semantic analysis, the current environment can be confirmed, and positioning and mapping can be completed. Specifically, visual perception technology can be combined to detect the position and direction of semantic elements such as parking spaces, pillars, walls, arrows, roadblocks, speed bumps, lane lines, etc. in the parking lot, and the map trajectory can be obtained through dead reckoning through inertial navigation equipment and wheel speed sensors to construct the map.

[0245] The collection of the first global positioning data can be achieved through GNSS positioning or map positioning technology, and the map positioning technology can include SLAM positioning, high-precision map positioning, etc.

[0246] Among them, as an implementation method, in the starting state before the vehicle enters the parking lot, a first semantic map of a preset area is continuously constructed, and first global positioning data is collected at the same time; the first global positioning data is temporally associated with the corresponding trajectory points on the first semantic map.

[0247] Among them, the preset area range can be selected and set according to actual conditions, and the accuracy requirements of the positioning data can also be considered. For example, it can be set in a preset area near the parking lot, or in an area within a preset distance from the parking lot.

[0248] Among them, when collecting the first global positioning data, it is necessary to temporally associate the first global positioning data with the trajectory points of the second semantic map, that is, the semantic elements corresponding to the trajectory points in the constructed map are associated with the global positioning data collected at the current moment to ensure that the positioning data and the information of the map semantic elements are synchronized.

[0249] After detecting that the global positioning signal is lost, the first semantic map continues to be constructed until it is detected that the vehicle enters the parking lot and is powered off.

[0250] The construction of the first semantic map and the collection of the first global positioning data are used for subsequent map splicing and correction.

[0251] This embodiment uses the above solution to construct maps for parking lot entrances and exits, while also collecting positioning data before and after global positioning signal loss. The spliced ​​map is then corrected by combining map splicing, constraint rules, and map fusion optimization methods, thereby reducing map construction errors and improving the accuracy of parking lot map construction.

[0252] Reference Figure 8 , Figure 8 This is a flow chart of the fourth embodiment of the map generation method of the present invention. Figure 7 Based on the embodiment shown, in the above step S104, before splicing the first semantic map and the second semantic map to obtain the spliced ​​semantic map, the method further includes:

[0253] Step S1041, determining whether the accuracy of the second global positioning data meets a preset standard;

[0254] If the accuracy of the second global positioning data meets the preset standard, step S104 is executed: the first semantic map and the second semantic map are spliced.

[0255] Specifically, in this embodiment, when collecting the second global positioning data, the accuracy requirements of the positioning data can be considered. For example, it can be set in a preset area near the parking lot, or set within an area within a preset distance from the parking lot. With sufficient data support, higher accuracy requirements for the positioning data can be achieved.

[0256] In addition, if Figure 9 As shown, an embodiment of the present invention further provides a map generating device, the map generating device comprising:

[0257] an acquisition module, configured to, upon detecting that a vehicle enters a parking lot and is powered off, acquire the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle is powered off, wherein the first global positioning data is temporally associated with the first semantic map;

[0258] A construction module, configured to construct a second semantic map based on the position and heading information of the vehicle at the time of power-off when detecting that the vehicle starts to leave the parking lot;

[0259] a collection and association module, configured to collect second global positioning data after detecting that the global positioning signal is restored, and temporally associate the second global positioning data with the second semantic map;

[0260] a splicing module, configured to splice the first semantic map and the second semantic map based on the first global positioning data and the second global positioning data to obtain a spliced ​​semantic map;

[0261] Furthermore, the device also includes:

[0262] A correction module is used to correct the spliced ​​semantic map based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method.

[0263] For the principle and implementation process of map generation in this embodiment, please refer to the above embodiments and will not be repeated here.

[0264] In addition, an embodiment of the present invention also proposes a terminal device, which includes a memory, a processor, and a map generation program stored in the memory and runnable on the processor. When the map generation program is executed by the processor, the steps of the map generation method described in the above embodiment are implemented.

[0265] Since all the technical solutions of all the aforementioned embodiments are adopted when this map generation program is executed by the processor, it has at least all the beneficial effects brought about by all the technical solutions of all the aforementioned embodiments, which will not be described one by one here.

[0266] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which a map generation program is stored. When the map generation program is executed by a processor, the steps of the map generation method described in the above embodiment are implemented.

[0267] Since all the technical solutions of all the aforementioned embodiments are adopted when this map generation program is executed by the processor, it has at least all the beneficial effects brought about by all the technical solutions of all the aforementioned embodiments, which will not be described one by one here.

[0268] Compared with the existing technology, the map generation method, apparatus, terminal device and storage medium proposed in the embodiments of the present invention obtain the position and heading information of the vehicle at the time of power-off, as well as the first semantic map constructed and the first global positioning data collected before the vehicle is powered off, when detecting that the vehicle enters the parking lot and powers off. The first global positioning data is temporally associated with the first semantic map, and the global positioning signal in the parking lot is lost. When it is detected that the vehicle starts to leave the parking lot, a second semantic map is constructed based on the position and heading information of the vehicle at the time of power-off. After detecting that the global positioning signal is restored, the second global positioning data is collected and the second global positioning data is temporally associated with the second semantic map. Based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map. Therefore, by constructing maps for entering and exiting the parking lot, and collecting positioning data before and after the global positioning signal is lost and restored, map splicing is performed based on the first global positioning data and the second global positioning data, thereby reducing map construction errors and improving the accuracy of parking lot map construction; further, based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected, thereby combining map splicing, constraint rules and map fusion optimization method to correct the spliced ​​map, further reducing map construction errors and improving the accuracy of parking lot map construction.

[0269] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or method that includes a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or method. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, article, or method that includes the element.

[0270] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0271] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, controlled terminal, or network device, etc.) to execute the method of each embodiment of the present invention.

[0272] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A map generation method, characterized in that: The method comprises the following steps: Upon detecting that a vehicle has entered a parking lot and powered off, obtaining the vehicle's location and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle was powered off, wherein the first global positioning data is temporally associated with the first semantic map. The global positioning signal is lost within the parking lot. The first semantic map includes at least the position and direction of semantic elements and a vehicle movement trajectory obtained via dead reckoning. When a vehicle is detected to have started to leave the parking lot, a second semantic map is constructed based on the vehicle's position and heading information at the time of power-off, the second semantic map including at least the position and direction of semantic elements and the vehicle's movement trajectory obtained by dead reckoning. After detecting that the global positioning signal has been restored, collecting second global positioning data, and temporally associating the second global positioning data with the second semantic map; Based on the first global positioning data and the second global positioning data, the first semantic map and the second semantic map are spliced ​​to obtain a spliced ​​semantic map.

2. The method according to claim 1, characterized in that The method further comprises: Based on the first global positioning data, the second global positioning data and preset constraint rules, and in combination with a preset map fusion optimization method, the spliced ​​semantic map is corrected.

3. The method according to claim 1, characterized in that Before the step of obtaining the vehicle's position and heading information at the time of power-off, as well as the first semantic map constructed and the first global positioning data collected before the vehicle is powered off, upon detecting that the vehicle enters the parking lot and powers off, the method further includes: In a startup state before the vehicle enters the parking lot, continuously constructing a first semantic map of a preset area and collecting first global positioning data; Associating the first global positioning data with corresponding track points on the first semantic map in time; After detecting that the global positioning signal is lost, the first semantic map continues to be constructed until it is detected that the vehicle enters the parking lot and is powered off.

4. The method according to claim 1, wherein Before the step of splicing the first semantic map and the second semantic map to obtain a spliced ​​semantic map, the following step further comprises: Determining whether the accuracy of the second global positioning data meets a preset standard; If the accuracy of the second global positioning data reaches a preset standard, the step of splicing the first semantic map and the second semantic map is performed.

5. The method according to claim 2, characterized in that The step of correcting the spliced ​​semantic map based on the first global positioning data, the second global positioning data, and the preset constraint rules, and in combination with a preset map fusion optimization method, includes: Obtaining an overlapping area between the first semantic map and the second semantic map in the spliced ​​semantic map, and using matching semantic elements in the overlapping area as mutual constraints; and / or using position and heading changes of associated elements in the first semantic map as constraints, and using position and heading changes of associated elements in the second semantic map as constraints; Based on the first global positioning data and the second global positioning data, acquiring corresponding constrained positioning data; Based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method, the positions and directions of all semantic elements in the spliced ​​semantic map are obtained; Based on the solved positions and directions of all semantic elements in the spliced ​​semantic map, the spliced ​​semantic map is corrected.

6. The method according to claim 5, characterized in that Before the step of using the position and heading changes of the elements associated in the first semantic map as constraints and the step of using the position and heading changes of the elements associated in the second semantic map as constraints, the method further includes: Obtaining the associated semantic elements in the first semantic map, and obtaining the associated semantic elements in the second semantic map; The element position and heading change of the associated semantic element in the first semantic map is obtained based on the first global positioning data, and the element position and heading change of the associated semantic element in the second semantic map is obtained based on the second global positioning data.

7. The method according to claim 5, characterized in that The step of obtaining the positions and directions of all semantic elements in the spliced ​​semantic map based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method includes: Solving the position error and / or height error between corresponding semantic elements in the spliced ​​semantic map based on the corresponding constrained positioning data and in combination with a preset map fusion optimization method; Based on the solved pose error and / or height error, the positions and directions of all semantic elements in the spliced ​​semantic map are solved.

8. A map generating device, characterized in that: The map generating device comprises: an acquisition module, configured to, upon detecting that a vehicle has entered a parking lot and powered off, acquire the vehicle's position and heading information at the time of power-off, as well as a first semantic map constructed and first global positioning data collected before the vehicle was powered off, wherein the first global positioning data is temporally associated with the first semantic map, and the first semantic map includes at least the position and direction of semantic elements and a vehicle movement trajectory obtained via dead reckoning; a construction module, configured to construct a second semantic map based on the vehicle's position and heading information at the time of power-off when detecting that the vehicle has started to leave the parking lot, the second semantic map including at least the position and direction of semantic elements and a vehicle movement trajectory obtained by dead reckoning; a collection and association module, configured to collect second global positioning data after detecting that the global positioning signal is restored, and temporally associate the second global positioning data with the second semantic map; The splicing module is configured to splice the first semantic map and the second semantic map based on the first global positioning data and the second global positioning data to obtain a spliced ​​semantic map.

9. A terminal device, characterized in that: The terminal device includes a memory, a processor, and a map generation program stored in the memory and executable on the processor. When the map generation program is executed by the processor, the steps of the map generation method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a map generation program, which, when executed by a processor, implements the steps of the map generation method according to any one of claims 1 to 7.

Images