Motor vehicle parking lot scene map construction method and device and computer readable storage medium

Abstract

The embodiment of the invention provides a motor vehicle parking lot scene map construction method and device and a computer readable storage medium, and the method comprises the steps: constructing a global map coordinate system with a preset starting point of a parking lot as a coordinate origin, and controlling a motor vehicle to drive from the preset starting point to a preset terminal point; in the driving process of the motor vehicle, generating a panoramic top view based on video images collected by each vehicle-mounted camera; obtaining the current position coordinates of the motor vehicle in a predetermined period, and constructing a local scene map of the motor vehicle based on the physical coordinates of a predetermined target in the panoramic top view; mapping the local scene map constructed in each period into the global map coordinate system according to the current position coordinate obtained in the corresponding period to obtain an initial global scene map; and eliminating repeated physical coordinates of the same predetermined target in the initial global scene map to obtain a target global scene map. The scene map constructed by the embodiment of the invention can be used for assisting in parking space selection and assisting in realizing path planning.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of motor vehicle mapping, and in particular, to a method, device, and computer-readable storage medium for constructing a map of a motor vehicle parking lot scene. Background Art

[0002] As the intelligence level of motor vehicles gradually increases, motor vehicles begin to be equipped with an automatic parking system. To achieve accurate path planning and parking navigation, existing automatic parking systems usually need to construct a map of the parking lot scene from a preset starting point to a preset ending point of the motor vehicle in the parking lot.

[0003] A conventional method for constructing a map of a parking lot scene for an automatic parking system mainly uses image data captured by an in-vehicle camera and visual SLAM technology. During the map construction process, it is necessary to extract semantic information of specific targets such as parking spaces, lane lines, and columns in the parking lot, and integrate the semantic information into the constructed map of the parking lot scene.

[0004] However, the inventor found in specific implementations that after extracting the semantic information of the specific targets, the above method directly integrates the semantic information into the map of the parking lot scene without further processing, resulting in the relevant semantic information being difficult to accurately reflect the exact positions of the respective specific targets, and the constructed map of the parking lot scene cannot be used to guide the selection of parking spaces and achieve path planning. Eventually, the constructed map of the parking lot scene can only achieve automatic parking tasks with fixed trajectories and cannot flexibly achieve automatic parking in dynamic scenarios according to changes in the implementation scenario. Summary of the Invention

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a method for constructing a map of a motor vehicle parking lot scene, and the constructed scene map can be used to assist in the selection of parking spaces and assist in achieving path planning.

[0006] The technical problem to be further solved by the embodiments of the present invention is to provide a device for constructing a map of a motor vehicle parking lot scene, and the constructed scene map can be used to assist in the selection of parking spaces and assist in achieving path planning.

[0007] The technical problem to be further solved by the embodiments of the present invention is to provide a computer-readable storage medium that can store a computer program for constructing a scene map that can be used to assist in the selection of parking spaces and assist in achieving path planning.

[0008] To solve the above technical problems, the embodiments of the present invention provide the following technical solutions: A method for constructing a map of a motor vehicle parking lot scene includes the following steps: Construct a global map coordinate system with the preset starting point of the parking lot as the coordinate origin, and control the motor vehicle to travel from the preset starting point to the preset ending point according to the preset driving path; During the driving process of the motor vehicle, control each on-vehicle camera of the motor vehicle to collect video images of the surrounding environment of the motor vehicle in real time, and generate a panoramic top view based on the video images collected by each on-vehicle camera; Obtain the current position coordinates of the motor vehicle in the global map coordinate system at a predetermined period and construct a local scene map based on the current position coordinates. The construction process of the local scene map includes: extracting the semantic information of predetermined targets in the parking lot from the panoramic top view, processing the semantic information to obtain the physical coordinates of the predetermined targets with actual physical lengths, and constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates in the panoramic top view. The predetermined targets at least include parking spaces, lane lines, and columns; Map the local scene maps constructed in each period to the global map coordinate system according to the current position coordinates obtained in the corresponding period to obtain an initial global scene map; and Exclude the repeated physical coordinates of the same predetermined target in the initial global scene map to obtain the target global scene map of the parking lot.

[0009] Further, the semantic information of the predetermined target refers to the pixel regions of the parking space lines, lane lines, and columns respectively obtained by performing semantic segmentation on the panoramic top view using a predetermined deep network model. The processing of the semantic information to obtain the physical coordinates of the predetermined target with actual physical lengths specifically includes: Perform morphological operations on the pixel regions of the parking space lines, lane lines, and columns respectively to fill the pixel holes in the pixel regions; Perform image processing on the pixel regions of the parking space lines, lane lines, and columns after the morphological operations to obtain the pixel coordinates of each parking space, each lane line, and each column in the panoramic top view respectively; and Convert the pixel coordinates of each parking space, each lane line, and each column in the panoramic top view into the corresponding physical coordinates.

[0010] Further, the pixel coordinates of the parking space are obtained through the following steps: successively processing the pixel region of the parking space line after morphological operations by using a predetermined line detection algorithm model and a predetermined line clustering algorithm model to obtain several parking space lines, and constructing the parking space and its pixel coordinates in the panoramic top view based on the relative position relationship of each parking space line.

[0011] Further, the pixel coordinates of the lane line are obtained through the following steps: successively processing the pixel region of the lane line after morphological operations by using a predetermined line detection algorithm model and a predetermined line clustering algorithm model to obtain several suspected lines, and excluding the interference lines with lengths less than a preset length threshold and horizontal distribution from each suspected line to obtain each lane line and its pixel coordinates in the panoramic top view.

[0012] Further, the pixel coordinates of the column are obtained through the following steps: calculating the shortest pixel distance between each pixel point in the pixel region of the column and the parking space line and the lane line, removing the pixel points in the pixel region of the column with the shortest pixel distance greater than a preset pixel distance, successively processing the remaining pixel points screened in the pixel region of the column by using a predetermined line detection algorithm model and a predetermined line clustering algorithm model to obtain the connection line between the column and the parking lot ground, and constructing the column and its pixel coordinates in the panoramic top view based on the connection line.

[0013] Further, the exclusion of the duplicate physical coordinates of the same predetermined target in the initial global scene map specifically includes: Excluding the remaining physical coordinates in the global scene map that belong to the same parking space and except for the latest physical coordinates; Excluding the remaining physical coordinates in the global scene map that belong to the same column and except for the latest physical coordinates; and Merging the physical coordinates in the global scene map that belong to the same lane line by using a predetermined line clustering algorithm model.

[0014] Further, after excluding the duplicate physical coordinates of the same predetermined target in the initial global scene map, an image constraint is also imposed on each parking space in the initial global scene map to obtain the target global scene map, and the image constraint includes the same size constraint and the lane line alignment constraint.

[0015] Further, after constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates, the usage status of the parking spaces in the local scene map is marked, and the usage status of the parking spaces is mapped to the target global scene map. Moreover, each time the target global scene map is used for assisted driving, the usage status of the parking spaces in the target global scene map is also updated. The method for marking the usage status of the parking spaces in the local scene map and updating the usage status of the parking spaces in the target global scene map includes: Identifying obstacle vehicles around the motor vehicle from each of the video images based on a predetermined vehicle recognition model; Calculating coordinate data of the obstacle vehicles in the global map coordinate system; and Marking or updating the usage status of the parking spaces according to the coordinate data of the obstacle vehicles.

[0016] On the other hand, to solve the above technical problems, an embodiment of the present invention provides the following technical solution: A device for constructing a motor vehicle parking lot scene map is connected to the driving system of the motor vehicle and each on-vehicle camera. The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for constructing a motor vehicle parking lot scene map as described in any one of the above.

[0017] On yet another hand, to solve the above technical problems, an embodiment of the present invention provides the following technical solution: A computer-readable storage medium includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing a motor vehicle parking lot scene map as described in any one of the above.

[0018] After adopting the above technical solution, the embodiment of the present invention has at least the following beneficial effects: After extracting the semantic information of predetermined targets such as parking spaces, lane lines, and columns in the parking lot from the panoramic top view, the embodiment of the present invention further processes the semantic information to obtain the physical coordinates of the predetermined targets with actual physical lengths, and constructs a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates. As a result, the parking spaces, lane lines, and columns in the constructed local scene map and the initial global scene map and target global scene map further generated based on the local scene map all contain corresponding physical coordinates. Therefore, when the automatic parking system uses the finally constructed target global scene map, it can use the physical coordinates of the parking spaces, lane lines, and columns therein for dynamic selection of parking spaces and dynamic planning of parking paths, which is beneficial to improving the efficiency and success rate of automatic parking. Description of the Drawings

[0019] Figure 1 This is a flowchart of the steps of an alternative embodiment of the method for constructing a scene map of a motor vehicle parking lot according to the present invention.

[0020] Figure 2 This is a schematic block diagram of an alternative embodiment of the device for constructing a scene map of a motor vehicle parking lot according to the present invention.

[0021] Figure 3 This is a functional module diagram of an alternative embodiment of the device for constructing a scene map of a motor vehicle parking lot according to the present invention. Detailed implementation manners

[0022] The following further describes the present application in detail with reference to the accompanying drawings and specific embodiments. It should be understood that the following illustrative embodiments and descriptions are only used to explain the present invention and are not intended to limit the present invention. Moreover, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0023] As Figure 1 shown, an alternative embodiment of the present invention provides a method for constructing a scene map of a motor vehicle parking lot, including the following steps: S1: Construct a global map coordinate system with a preset starting point of the parking lot as the coordinate origin, and control the motor vehicle to travel from the preset starting point to a preset ending point along a preset driving path; S2: During the driving process of the motor vehicle, control each on-vehicle camera 1 of the motor vehicle to collect video images of the surrounding environment of the motor vehicle in real time, and generate a panoramic top view based on the video images collected by each on-vehicle camera; S3: Obtain the current position coordinates of the motor vehicle in the global map coordinate system at a predetermined period and construct a local scene map based on the current position coordinates. The construction process of the local scene map includes: extracting semantic information of a predetermined target in the parking lot from the panoramic top view, processing the semantic information to obtain physical coordinates of the predetermined target with an actual physical length, and constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates in the panoramic top view. The predetermined target at least includes a parking space, a lane line, and a column; S4: Map the local scene maps constructed in each period to the global map coordinate system according to the current position coordinates obtained in the corresponding period to obtain an initial global scene map; and S5: Exclude the repeated physical coordinates of the same predetermined target in the initial global scene map to obtain a target global scene map of the parking lot.

[0024] In an embodiment of the present invention, after extracting semantic information of predetermined targets such as parking spaces, lane lines, and columns in a parking lot from a panoramic top view, the semantic information is further processed to obtain physical coordinates of the predetermined targets with actual physical lengths, and a local scene map of the motor vehicle at the current position coordinates is constructed based on the physical coordinates. As a result, the parking spaces, lane lines, and columns in the constructed local scene map, the initially generated global scene map, and the target global scene map further generated based on the local scene map all include corresponding physical coordinates. Therefore, when the final constructed target global scene map is used in the automatic parking system, the physical coordinates of the parking spaces, lane lines, and columns therein can be used for dynamic selection of parking spaces and dynamic planning of parking paths, which is beneficial to improving the efficiency and success rate of automatic parking.

[0025] In specific implementation, in step S1, when constructing the global map coordinate system, a preset starting point (when the motor vehicle starts to drive, the geometric center of the motor vehicle is also located at the preset starting point) is used as the coordinate origin, the head direction is the positive direction of the y-axis, and the right direction perpendicular to the y-axis is the positive direction of the x-axis; in addition, it can be the driving of the motor vehicle by manual driving or the automatic control of the driving of the motor vehicle by an intelligent driving system; in step S2, when forming a panoramic top view based on the video images collected by each vehicle-mounted camera 1, the pre-calibrated external camera parameters of the vehicle-mounted camera 1 are required. After converting each frame of the four-way video images into a top view through perspective transformation, the images in the top view are stitched and fused to form the panoramic top view; in step S3, when constructing the local scene map, a two-dimensional coordinate system is constructed with the geometric center of the vehicle as the origin, the positive direction of the x-axis to the right, and the positive direction of the y-axis upward. Then, the detected parking spaces, lane lines, and columns are marked in the two-dimensional coordinate system according to the physical coordinates, so as to obtain a local scene map containing semantics; and in step S4, in addition, the period can be every predetermined time length (for example: 5 s) during the uniform driving of the motor vehicle or every driving distance (for example: 50 cm) of the motor vehicle. This period parameter can be obtained from the odometer of the motor vehicle.

[0026] In an alternative embodiment of the present invention, the semantic information of the predetermined target refers to the pixel regions of the parking space lines, lane lines, and columns respectively obtained by performing semantic segmentation on the panoramic top view using a predetermined deep network model. The processing of the semantic information to obtain the physical coordinates of the predetermined target with actual physical lengths specifically includes: Performing morphological operations on the pixel regions of the parking space lines, the pixel regions of the lane lines, and the pixel regions of the columns respectively to fill the pixel holes in the pixel regions; Perform image processing on the pixel regions of the parking space lines, the lane lines, and the columns after morphological operations to respectively obtain the pixel coordinates of each parking space, the pixel coordinates of each lane line, and the pixel coordinates of each column in the panoramic top view; and Convert the pixel coordinates of each parking space, the pixel coordinates of each lane line, and the pixel coordinates of each column in the panoramic top view into the corresponding physical coordinates respectively.

[0027] In this embodiment, by using a predetermined deep network model to perform semantic segmentation on the panoramic top view, the category pixels pre-standardized in the panoramic top view can be segmented, so as to distinguish the pixel regions of the parking space lines, the lane lines, and the columns; when further processing the semantic information, morphological operations are performed on the pixel regions to fill the pixel holes in the pixel regions, thereby improving the robustness of subsequent processing and avoiding data errors or failures in subsequent image processing due to pixel holes; further, by performing image processing on the pixel regions of the parking space lines, the lane lines, and the columns, the pixel coordinates of the parking spaces, the pixel coordinates of each lane line, and the pixel coordinates of each column can be obtained. Finally, combined with the external parameters of the vehicle-mounted camera 1, the corresponding pixel coordinates can be converted into the corresponding physical coordinates. The overall process is simple and the processing efficiency is high.

[0028] In specific implementation, the predetermined deep network model can adopt network models such as U-net, DeepLab, or SegNe. In this embodiment of the present invention, the U-net network model is adopted. When pre-training the U-net network model, first crop the training pictures to a unified size and manually label the categories in the cropped images. Adjust the number of channels in the last layer of the U-net network model to adjust the number of categories for the required semantic segmentation. After obtaining the trained network model, this network model can be used to perform semantic segmentation on the panoramic top view; in addition, it can be understood that when performing semantic segmentation, the lane ground markings and the drivable free area markings in the panoramic top view will also be extracted, and the above markings will also participate in constructing the final target global scene map. When the constructed target global scene map is applied to the automatic parking system, the lane ground markings and the drivable free area markings therein can assist in path planning.

[0029] In an alternative embodiment of the present invention, the pixel coordinates of the parking space are obtained through the following steps: successively using a predetermined line detection algorithm model and a predetermined line clustering algorithm model to process the pixel region of the parking space line after morphological operations to obtain several parking space lines, and constructing the parking space and its pixel coordinates in the panoramic top view based on the relative position relationship of each parking space line. In this embodiment, when processing the pixel region of the parking space line, line detection and line clustering are successively performed, so that the interference lines can be greatly reduced and the data processing amount can be reduced; after determining the parking space lines, since the intersection of two parking space lines is a corner point of the parking space, and according to the relative position relationship of each parking space line, several parking space lines can be quickly surrounded to generate the corresponding parking space, and after determining the parking space, its pixel coordinates in the panoramic top view can also be determined.

[0030] In an alternative embodiment of the present invention, the pixel coordinates of the lane line are obtained through the following steps: successively using a predetermined line detection algorithm model and a predetermined line clustering algorithm model to process the pixel region of the lane line after morphological operations to obtain several suspected lines, and excluding the interference lines with a length less than a preset length threshold and horizontal distribution from each suspected line to obtain each lane line and its pixel coordinates in the panoramic top view. In this embodiment, for the lane line, it is also usually a straight line in the actual scene. Based on this, line detection and line clustering are also successively performed, so that the interference lines can be greatly reduced and the data processing amount can be reduced. After determining several suspected lines, since the lane line usually has a certain length and is not a horizontally distributed line, therefore, after excluding such interference lines, the remaining suspected lines are determined as lane lines, and after determining the lane line, its pixel coordinates in the panoramic top view can also be determined.

[0031] In an alternative embodiment of the present invention, the pixel coordinates of the column are obtained through the following steps: calculate the shortest pixel distance between each pixel point in the pixel area of the column and the parking space line and the lane line, remove the pixel points in the pixel area of the column where the shortest pixel distance is greater than the preset pixel distance, and sequentially use a predetermined line detection algorithm model and a predetermined line clustering algorithm model to process the remaining pixel points in the pixel area of the column to obtain the connection line between the column and the parking lot ground, and construct the column and its pixel coordinates based on the connection line in the panoramic top view. In this embodiment, for the column, since the connection line between the column and the parking lot ground and the parking space line and the lane line are all on the ground, therefore, the shortest pixel distance between the pixel points on the connection line and the parking space line and the lane line usually does not exceed the preset pixel distance. Based on this, the pixel points located on the connection line are filtered out, and then line detection and line clustering are also used to obtain the connection line and its pixel coordinates. After determining the connection line and its pixel coordinates, based on multiple different connection lines belonging to the same column, the column and its pixel coordinates can be determined.

[0032] In specific implementation, the predetermined line detection algorithm model mentioned in the foregoing embodiments may be the Hough line detection, LSD line detection algorithm model, etc. In the embodiment of the present invention, the Hough line detection algorithm model is selected, and for the line clustering algorithm model, it may be the K-means algorithm. The clustering method adopted in the embodiment of the present invention is as follows: The first step: sort the y values of the starting point pixel coordinates of the line segments from large to small; The second step: sequentially calculate the lengths of each line segment, and remove the line segments with lengths less than the preset length; The third step: sequentially calculate the slopes of the remaining line segments, compare two line segments with a slope difference less than the preset difference threshold. If the relative distance between the starting point or the ending point of these two line segments is less than the preset distance, then connect the starting point of the first line segment and the ending point of the second line segment to form a new line segment; The fourth step: loop the third step, traverse all line segments, and the finally obtained line segments are the clustered line segments.

[0033] In an alternative embodiment of the present invention, step S4 specifically includes: Exclude the remaining physical coordinates in the global scene map that belong to the same parking space and except for the latest physical coordinates; Exclude the remaining physical coordinates in the global scene map that belong to the same column and except for the latest physical coordinates; and Merge the physical coordinates in the global scene map that belong to the same lane line using a predetermined line clustering algorithm model.

[0034] In this embodiment, for the physical coordinates of the parking spaces and the columns, only the latest physical coordinates (i.e., the physical coordinates obtained by image calculation for the last time) are retained. For each line segment of the same lane line in the global scene map, a predetermined straight line clustering algorithm model is used for merging (which can be merged into one or a smaller number of lines), thereby reducing the amount of data processing.

[0035] In an alternative embodiment of the present invention, after excluding the physical coordinates of the same predetermined target that are repeated in the initial global scene map, an image constraint is further imposed on each parking space in the initial global scene map to obtain the target global scene map. The image constraint includes the same size constraint and the lane line alignment constraint. Since the length and width dimensions of the parking spaces in the same parking lot should be kept consistent, and at the same time, the sides of adjacent or side-by-side parking spaces should be aligned with the lane lines. Based on this, in this embodiment, the same size constraint and the lane line alignment constraint are further imposed on each parking space in the global scene map, thereby improving the position accuracy of the parking spaces.

[0036] In an alternative embodiment of the present invention, after constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates, the usage status of the parking spaces in the local scene map is further marked, and the usage status of the parking spaces is mapped to the target global scene map. Moreover, each time the target global scene map is used for assisted driving, the usage status of the parking spaces in the target global scene map is also updated. Among them, the method for marking the usage status of the parking spaces in the local scene map and updating the usage status of the parking spaces in the target global scene map includes: Based on a predetermined vehicle recognition model (for example: the yolov5 algorithm model), obstacle vehicles around the motor vehicle are recognized from each of the video images; Calculating the coordinate data of the obstacle vehicles in the global map coordinate system; and Marking or updating the usage status of the parking spaces according to the coordinate data of the obstacle vehicles.

[0037] In this embodiment, during the process of constructing the local scene map and during the subsequent use of the target global scene map, the usage status of the parking spaces is further marked or updated, so that when the target global scene map is applied, it is convenient for the system to accurately determine whether a parking space is occupied in the short term, and thus flexibly select an idle parking space for parking.

[0038] On the other hand, an alternative embodiment of the present invention provides a device 3 for constructing a motor vehicle parking lot scene map, which is connected to the driving system 5 of the motor vehicle and each on-vehicle camera 1. The device 3 includes a processor 30, a memory 32, and a computer program stored in the memory 32 and configured to be executed by the processor 30. When the processor 30 executes the computer program, the method for constructing a motor vehicle parking lot scene map as described in any one of the above is implemented.

[0039] Exemplarily, the computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory 32 and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the device 3 for constructing a motor vehicle parking lot scene map. For example, the computer program can be divided into Figure 3 function modules in the device 3 for constructing a motor vehicle parking lot scene map, wherein a motor vehicle control module 41, an image processing module 42, a local map construction module 43, an initial global map construction module 44, and a target global map construction module 45 respectively execute the above steps S1 - S5.

[0040] The device 3 for constructing a motor vehicle parking lot scene map can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The device 3 for constructing a motor vehicle parking lot scene map may include, but is not limited to, a processor 30 and a memory 32. Those skilled in the art can understand that the schematic diagram is only an example of the device 3 for constructing a motor vehicle parking lot scene map, and does not constitute a limitation on the device 3 for constructing a motor vehicle parking lot scene map. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the device 3 for constructing a motor vehicle parking lot scene map may further include input / output devices, network access devices, a bus, etc.

[0041] The processor 30 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor 30 is the control center of the motor vehicle parking lot scene map construction device 3, and connects various parts of the entire motor vehicle parking lot scene map construction device 3 through various interfaces and lines.

[0042] The memory 32 can be used to store the computer programs and / or modules. The processor 30 realizes various functions of the motor vehicle parking lot scene map construction device 3 by running or executing the computer programs and / or modules stored in the memory 32, and by calling the data stored in the memory 32. The memory 32 may mainly include a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a graphic recognition function, a graphic overlay function, etc.); the data storage area can store data created according to the use of the control device (such as graphic data, etc.). In addition, the memory 32 may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0043] If the functions described in the embodiments of the present invention are implemented in the form of software function modules or units and sold or used as independent products, they can be stored in a computer-readable storage medium that can be read by a computing device. Based on such an understanding, to implement all or part of the processes in the above-described method embodiments, the present invention can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor 30, the steps of the above-described method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0044] In another aspect, an optional embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the method for constructing a scene map of a motor vehicle parking lot as described in any one of the above.

[0045] The various embodiments in this specification are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0046] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit and scope protected by the claims of the present invention. All of these fall within the protection scope of the present invention.

Claims

1. A method for constructing a motor vehicle parking lot scene map, characterized in that: The method comprises the following steps: Constructing a global map coordinate system with a preset starting point of the parking lot as the coordinate origin, and controlling the motor vehicle to travel from the preset starting point to the preset end point along a preset driving path; During the driving of the motor vehicle, controlling each vehicle-mounted camera of the motor vehicle to collect video images of the surrounding environment of the motor vehicle in real time, and generating a panoramic bird's-eye view based on the video images collected by each vehicle-mounted camera; The current position coordinates of the motor vehicle in the global map coordinate system are obtained at a predetermined period and a local scene map based on the current position coordinates is constructed, wherein the construction process of the local scene map includes: extracting semantic information of a predetermined target in the parking lot from the panoramic bird's-eye view, processing the semantic information to obtain physical coordinates of the predetermined target with an actual physical length, and constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates, wherein the predetermined target includes at least a parking space, a lane line, and a pillar; Mapping the local scene map constructed in each period according to the current position coordinates obtained in the corresponding period into the global map coordinate system to obtain an initial global scene map; and The physical coordinates repeated for the same predetermined target in the initial global scene map are excluded to obtain a target global scene map of the parking lot.

2. The method for constructing a motor vehicle parking lot scene map according to claim 1, characterized in that: The semantic information of the predetermined target refers to the pixel area of ​​the parking space line, the pixel area of ​​the lane line and the pixel area of ​​the pillar obtained by semantically segmenting the panoramic bird's-eye view using a predetermined deep network model. The processing of the semantic information to obtain the physical coordinates of the predetermined target with an actual physical length specifically includes: Performing morphological operations on the pixel area of ​​the parking space line, the pixel area of ​​the lane line, and the pixel area of ​​the pillar respectively to fill the pixel holes in the pixel area; Performing image processing on the pixel areas of the parking space lines, the pixel areas of the lane lines, and the pixel areas of the pillars after the morphological operation to respectively obtain pixel coordinates of each parking space, pixel coordinates of each lane line, and pixel coordinates of each pillar in the panoramic bird's-eye view; and The pixel coordinates of each parking space, the pixel coordinates of each lane line, and the pixel coordinates of each pillar in the panoramic bird's-eye view are respectively converted into the corresponding physical coordinates.

3. The method for constructing a motor vehicle parking lot scene map according to claim 2, characterized in that: The pixel coordinates of the parking space are obtained by the following steps: a predetermined straight line detection algorithm model and a predetermined straight line clustering algorithm model are used in sequence to process the pixel area of ​​the parking space line after the morphological operation to obtain a plurality of parking space lines, and a parking space and its pixel coordinates are constructed in the panoramic bird's-eye view based on the relative position relationship of each parking space line.

4. The method for constructing a motor vehicle parking lot scene map according to claim 2, characterized in that: The pixel coordinates of the lane line are obtained by the following steps: a predetermined straight line detection algorithm model and a predetermined straight line clustering algorithm model are used in sequence to process the pixel area of ​​the lane line after morphological operation to obtain a number of suspected lines, and interference lines with a length less than a preset length threshold and horizontal distribution are excluded from each of the suspected lines to obtain each lane line and its pixel coordinates in the panoramic overhead view.

5. The method for constructing a motor vehicle parking lot scene map according to claim 2, 3 or 4, characterized in that: The pixel coordinates of the column are obtained by the following steps: calculating the shortest pixel distance between each pixel point in the pixel area of ​​the column and the parking space line and the lane line, removing the pixel points in the pixel area of ​​the column whose shortest pixel distance is greater than a preset pixel distance, and sequentially using a predetermined straight line detection algorithm model and a predetermined straight line clustering algorithm model to process the remaining pixel points in the pixel area of ​​the column to obtain a connecting line between the column and the parking lot ground, and constructing the column and its pixel coordinates in the panoramic bird's-eye view based on the connecting line.

6. The method for constructing a motor vehicle parking lot scene map according to claim 1, characterized in that: The physical coordinates for excluding duplication of the same predetermined target in the initial global scene map specifically include: Excluding the remaining physical coordinates in the global scene map that belong to the same parking space except for the latest physical coordinates; Excluding the remaining physical coordinates in the global scene map that belong to the same column except for the latest physical coordinates; and The physical coordinates belonging to the same lane line in the global scene map are merged using a predetermined straight line clustering algorithm model.

7. The method for constructing a motor vehicle parking lot scene map according to claim 1 or 6, characterized in that: After excluding the repeated physical coordinates of the same predetermined target in the initial global scene map, image constraints are also performed on each parking space in the initial global scene map to obtain the target global scene map, and the image constraints include same size constraints and lane alignment constraints.

8. The method for constructing a motor vehicle parking lot scene map according to claim 1, characterized in that: After constructing a local scene map of the motor vehicle at the current position coordinates based on the physical coordinates, the use status of the parking space in the local scene map is also marked, and the use status of the parking space is mapped to the target global scene map, and each time the target global scene map is used for assisted driving, the use status of the parking space in the target global scene map is also updated, wherein the method for marking the use status of the parking space in the local scene map and updating the use status of the parking space in the target global scene map includes: Identifying obstacle vehicles around the motor vehicle from each of the video images based on a predetermined vehicle recognition model; Calculating and obtaining coordinate data of the obstacle vehicle in the global map coordinate system; and The use status of the parking space is marked or updated according to the coordinate data of the obstacle vehicle.

9. A motor vehicle parking lot scene map construction device, connected to a motor vehicle driving system and each vehicle-mounted camera, characterized in that: The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for constructing a motor vehicle parking lot scene map as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the motor vehicle parking lot scene map construction method as described in any one of claims 1 to 8.

Images