Mapping correction system

Receive image information and driving data through four monocular cameras, build and correct two-dimensional or three-dimensional point cloud maps of parking lots, solve the problems of inaccurate positioning and large resource occupation in the existing technology, and improve the map of autonomous parking Establish efficiency and accuracy.

CN119043296APending Publication Date: 2024-11-29OTOBRITE ELECTRONICS INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310590328.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-24
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing autonomous parking technology in indoor parking lots is inaccurate due to poor GPS signal, resulting in inaccurate vehicle positioning. The system resources are occupied during the establishment of a three-dimensional point cloud map, which affects other functional operations. The image reconstruction effect is not ideal, making it easy to appear. Problems of duplicate features or poor splicing.

Method used

Four monocular cameras are used to receive continuous image information around the vehicle. The processing unit generates depth feature points through driving data information, constructs a two-dimensional or three-dimensional point cloud map, and matches and corrects the map in the overlapping area of ​​the driving trajectory, and removes dynamic features through semantic operations. points, save storage capacity and improve map accuracy.

Benefits of technology

It realizes saving map storage capacity in autonomous parking, reducing feature point matching operations during map construction, improving feature point junction effect of parking lot maps, and enhancing the accuracy and reliability of the map.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119043296A_ABST
    Figure CN119043296A_ABST
Patent Text Reader

Abstract

The invention discloses a mapping correction system which is mounted on a vehicle and comprises an image information receiving module for receiving continuous image information which is shot by four monocular cameras at least mounted in front of and behind the vehicle and on the left and right sides of the vehicle and covers all the surrounding environment of the vehicle; the processing unit is used for receiving driving data information and generating a driving track of the vehicle; the processing unit is used for acquiring continuous image information of the surrounding environment of the vehicle, generating feature points with depth information in each frame of the continuous image information of the surrounding environment of the vehicle when the vehicle is subjected to a mapping program, and constructing and storing a point cloud map after the feature points are matched with driving data information of the vehicle, when the processing unit judges that the driving track of the vehicle is within an overlapping range, the stored point cloud maps are matched and then spliced to generate a corrected point cloud map.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a system for autonomous parking map building and correction. Background Art

[0002] With the change of lifestyle, the autonomous parking function brings convenience to users in specific parking areas, such as residential building parking lots, workplace parking lots, or parking lots frequently used by users. However, the autonomous parking function comes from the map information of the parking area, and the positioning of the vehicle comes from the establishment of accurate maps. The common indoor parking areas are limited by the poor reception of the global satellite positioning system (GPS), so the realization of autonomous parking depends more on the establishment and identification of the field map by the vehicle itself. Simultaneous localization and mapping technology often uses three-dimensional point cloud maps. However, the system resources and storage capacity occupied by the three-dimensional map of the parking area by camera shooting are relatively large, which in turn affects the calculation or storage of other functions of the vehicle safety driving system. In addition, in the related technology, the reconstruction effect of the parking area map established only by the image is not ideal, especially due to the change of vehicle posture and the cumulative error caused by image ranging, resulting in repeated features or poor splicing of the parking area map. Summary of the invention

[0003] In order to solve the above technical problems, an object of the present invention is to provide a system for map building and correction.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] The present invention provides a mapping correction system installed on a vehicle, comprising:

[0006] An image information receiving module receives continuous image information covering all surroundings of the vehicle from at least four monocular cameras installed in front of, behind, and on both sides of the vehicle;

[0007] A processing unit receives driving data information to generate a driving trajectory of the vehicle; and when the vehicle performs a mapping procedure, the processing unit generates feature points with depth information in each frame of continuous image information of the vehicle's surrounding environment, and after matching the feature points with the driving data information of the vehicle, constructs and stores a point cloud map;

[0008] When the processing unit determines that the driving track of the vehicle is within an overlapping range, the stored point cloud maps are matched and spliced ​​to generate a revised point cloud map.

[0009] Wherein, the point cloud map includes a two-dimensional point cloud map and / or a three-dimensional point cloud map.

[0010] The driving data information is odometer information.

[0011] The processing unit is used to define a detection area of ​​continuous image information of the vehicle's surrounding environment; determine whether the detection area contains an identification target, wherein when the detection area includes the identification target, the processing unit constructs the two-dimensional point cloud map corresponding to the identification target; when the identification target is not included in the detection area, the processing unit constructs the three-dimensional point cloud map.

[0012] The first detection area and the second detection area are bird's-eye view areas projected on the ground.

[0013] The processing unit determines whether the driving track of the vehicle is within an overlapping range by comparing RTK information.

[0014] Wherein, the processing unit removes dynamic feature points in the three-dimensional point cloud map through semantic operations.

[0015] The present invention has the following advantages:

[0016] 1. When establishing a parking lot map during autonomous parking, since the system of the present invention receives continuous image frames of the surrounding environment of the vehicle captured by at least four monocular cameras, it can selectively construct a two-dimensional or three-dimensional point cloud map of the parking lot by identifying and recognizing targets in a specific detection area, thereby saving map storage capacity.

[0017] 2. By triggering the overlapping area of ​​the vehicle driving trajectory in the present invention, the operation of feature point matching that needs to be started at any time during the mapping process can be reduced, and the mapping error can be corrected by comparing the feature points of the surround view image. The joint effect of the feature points of the parking lot can be obtained by matching the two-dimensional or three-dimensional feature points of the surround view image. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a structural schematic diagram of an autonomous parking mapping system mounted on a vehicle according to the present invention;

[0019] Figure 2 It is a flow chart of a mapping and loopback joining method of the present invention;

[0020] Figure 3 is a schematic diagram of a camera of the present invention being arranged on a vehicle;

[0021] Figure 4 is a flow chart of generating feature points of continuous image frames of the present invention;

[0022] Figure 5 is a schematic diagram of the detection area of ​​the continuous image frame of the present invention;

[0023] Figure 6 It is a schematic diagram of the present invention for determining whether a detection area contains an identification target when a vehicle is moving;

[0024] Figure 7 It is a schematic diagram of the present invention when the driving trajectory of the vehicle is in the overlapping range. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention claimed for protection, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0026] In the description of the present invention, it is necessary to understand that the terms "upper" and "lower" etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0027] In the description of the present invention, “plurality” means two or more than two, unless otherwise clearly and specifically defined.

[0028] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two components or the interaction relationship between two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0029] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0030] Please refer to Figure 1 The present invention discloses a mapping correction system 100 installed on a vehicle, including a first camera 101, a second camera 102, a third camera 103 and a fourth camera 104, an image information receiving module 105, a vehicle information assembly interface 106, a processing unit 107 and a storage module 108.

[0031] In this system, the first camera 101 to the fourth camera 104 are usually single-purpose cameras, and the camera is usually composed of a housing, a lens, an image sensing circuit board, an image processing circuit board, a connector, a power line, an image output line or a control line. The lens of the camera is exposed to the outside through the lens slot of the housing, and generally has a fisheye lens or a wide-angle lens. The image sensing circuit board is provided with a photosensitive component corresponding to the lens, which converts the light signal from the lens into an electronic image signal. The image processing circuit board is provided with a processing module and a memory, the memory temporarily stores the electronic image signal, and the processing module processes the electronic image signal to generate a processed image. The connector electrically connects the image sensing circuit board and the image processing circuit board. The power line passes through the housing and is electrically connected to the image processing circuit board to provide power, and the image output line passes through the housing and is electrically connected to the image processing circuit board. The control line passes through the housing and is electrically connected to the image processing circuit board to input a control signal, and the image processing circuit board selectively outputs the processed image from the image output line according to the control signal. At the same time, these cameras 101~104 are usually installed on the outside of the vehicle to avoid being blocked by the vehicle's own field of view, and are installed on different sides of the vehicle. They are mainly used to capture continuous image frames of the vehicle's surrounding environment. In order to obtain better image quality, the continuous image frame format can be LVDS format information.

[0032] In the present invention, the image information receiving module 105 is connected to the first camera 101, the second camera 102, the third camera 103 and the fourth camera 104, and is used to receive continuous image information from these cameras shooting the surrounding environment of the vehicle. The image information receiving module 105 also has an image processing unit (ISP, Image Signal Processor), which can process functions such as lens correction, pixel correction, color interpolation, Bayer noise removal, white balance correction, color correction, gamma correction, color space conversion, etc. The image information receiving module 103 generally has an LVDS (Low Voltage Differential Signaling) or MIPI CSI transmission interface (not marked).

[0033] The vehicle information assembly interface 106 is connected to the vehicle CAN bus to receive driving data information from the vehicle. The driving data information of the vehicle includes, for example, vehicle speed, engine speed, steering angle, acceleration, gear position, etc., and may even include odometer information, i.e., sensing information from an inertial measurement unit (IMU) and wheel speed meter, RTK (Real Time Kinematic, real-time dynamic positioning technology) or GPS installed in the vehicle that can measure the moving distance of the vehicle.

[0034] The processing unit 107 is the main computing unit of the present invention, which has built-in optimized artificial intelligence and has the ability to perform semantic computing of images. Usually, the processor of the processing unit 107 is a DSP (digital signal processor). The DSP is suitable for performing various multiplication and addition operations (SOP: Sum of Products), such as: finite impulse response filter operation (FIR: Finite Impulse Response), infinite impulse response filter operation (IIR: Infinite Impulse Response), discrete Fourier transform (DFT: Discrete Fourier Transform), discrete cosine transform (DCT: Discrete Cosine Transform), dot product operation (Dot product), convolution operation (Convolution), and matrix polynomial evaluation operation, etc. The processing unit 107 is connected to the image information receiving module 105, and calculates the continuous image information converted by the image information receiving module 105. The processing unit 107 is also connected to the vehicle information assembly interface 106 to receive driving data information from the vehicle. Therefore, according to the continuous image information and driving data information of the vehicle's surrounding environment, when the autonomous parking mapping function is turned on, the continuous image frames of the four monocular cameras can be obtained to start building a point cloud map.

[0035] The main function of the storage module 108 is to store the continuous image information from the monocular camera and / or various driving data information, the vehicle driving trajectory generated by the processing unit 107, and the point cloud map after the autonomous parking map is completed. The storage module 108 can be a built-in integrated circuit memory or an external storage device such as an SSD or SD card. At the same time, the storage module 106 is connected to the processing unit 107 and can store data according to the information generated by the processing unit 107.

[0036] Please refer to Figures 1 to 7 , is an implementation example of the operation of a mapping correction system 100 according to the present invention.

[0037] Step S01, receiving continuous image information from the camera covering all vehicle surroundings and receiving driving data information to generate the vehicle's driving trajectory. In order to build a map of the parking lot, the user can start the map building mode after the vehicle drives to the parking lot. At the same time, the user must drive in the parking lot by himself. The system 100 of the present invention uses the image information receiving module 105 to receive the continuous image information of the front of the vehicle taken by the first camera 101 installed in front of the vehicle, and the continuous image information of the rear of the vehicle taken by the second camera 102 installed at the rear of the vehicle. At the same time, it also receives the continuous image information of the left and right sides of the vehicle taken by the third camera 103 and the fourth camera 104 installed on the left and right sides of the vehicle. Figure 3 As shown, the field of view of these cameras installed on the outside of the vehicle covers all continuous image information of the vehicle's surrounding environment, forming a continuous image information of the surrounding environment of the vehicle. Although the present invention uses at least four cameras to cover the continuous image information of all the vehicle's surrounding environment, it is not limited to this. Any configuration of multiple cameras that can form continuous image information covering all the vehicle's surrounding environment is regarded as the present invention. It is worth noting that the present invention only uses four cameras to cover the continuous image information of all the vehicle's surrounding environment. Figure 3 It means that the lens field of view of cameras 101 to 104 covers the surrounding environment of the vehicle, not the distance range of the continuous image information captured by the camera, that is, the setting of the camera of the present invention can form a panoramic field of view around the vehicle. However, if such a panoramic field of view of the vehicle is to further achieve more effective mapping and reduce the repeated calculation or matching misjudgment of feature points in the image information of two overlapping image areas (such as the overlapping area images of the first camera 101 and the third camera 103) in the subsequent steps, in the present invention, the continuous image information of the individual cameras can be further spliced ​​through the processing unit 107 and then the subsequent steps are calculated, that is, first spliced ​​into a continuous image frame of the whole field of view. In order to achieve the technical means of the present invention, these monocular cameras must have a good focal length, and the shooting within a distance of 20 meters appears clear, so that the user can completely collect the scene images of the parking area during the self-driving process, so that the vehicle has a corresponding complete two-dimensional or three-dimensional point cloud map for each frame of the image information of the vehicle's surrounding environment on the path to the parking area.

[0038] At the same time, the vehicle information assembly interface 106 is connected to the vehicle CAN bus to receive driving data information from the vehicle. The vehicle's driving data information includes, for example, vehicle speed, engine speed, steering angle, acceleration, gear position, etc., and may even include odometer information (i.e., from an inertial measurement unit (IMU), wheel speed meter, RTK (Real Time Kinematic, real-time dynamic positioning technology) installed in the vehicle), or GPS and other related sensing information that can measure the vehicle's moving distance. In the present invention, these vehicle driving data information is used as sensing information to measure the vehicle's moving distance and speed, so as to serve as actual distance information matching the continuous image information of the vehicle's surrounding environment. In addition, since the processing unit 107 is connected to the vehicle information assembly interface 106 and receives the driving data information, the vehicle's driving trajectory can be generated after calculation. Regarding the generation of the vehicle's driving trajectory, in this step, the processing unit 107 can establish the origin of the world coordinate system where the vehicle is located when the mapping mode is turned on. By calculating the driving data information, especially the odometer information, the driving trajectory with coordinate information relative to the world coordinate system origin where the vehicle is located can be gradually established during the vehicle's driving process.

[0039] Step S02, generating feature points with depth information in each frame of the continuous image information of the vehicle's surrounding environment. In this step, the processing unit 107 of the present invention extracts feature points from the continuous image information in step S01, and the feature points can be divided into two-dimensional feature points and three-dimensional feature points. Generally speaking, if a precise three-dimensional point cloud map is simply established, there is no need to additionally extract two-dimensional feature points of the continuous image information. However, in order to effectively reduce the huge amount of calculation and storage generated by the extraction of three-dimensional feature points during mapping and the subsequent storage and access of three-dimensional point cloud maps, the present invention will access two-dimensional point cloud maps and three-dimensional point cloud maps separately in the present invention to reduce the amount of calculation for opening three-dimensional feature point extraction for a long time during the mapping process.

[0040] Please refer to Figure 4 , which is a step in which the processing unit 107 of the present invention generates feature points with depth information in each frame of continuous image information of the vehicle's surrounding environment:

[0041] (1) Step S021, defining a first detection area 201 and a second detection area 202 of continuous image frames. The continuous image information received by the camera 103 on the left side of the vehicle and the camera 104 on the right side are processed by the image information receiving module 105 to form a continuous image frame, and then the continuous image information of the surrounding environment of the vehicle is defined by the processing unit 107. Figure 5, the detection area of ​​the present invention corresponds to the first detection area 201 on the left side of the vehicle and the second detection area 202 on the right side of the vehicle. In the present invention, usually the length (length in the direction parallel to the side of the vehicle) of the first detection area 201 and the second detection area 202 can be an effective range of 5 to 6 meters, and the width (depth in the direction perpendicular to the side of the vehicle) can be an effective range of 3 to 5 meters. However, the present invention is not limited thereto, and the effective detection range of the first detection area 201 and the second detection area 202 can be increased according to the improvement of the image resolution of the monocular camera. In this embodiment, the calculation of the first detection area 201 and the second detection area 202 by the processing unit 107 can be framed by a three-dimensional spatial range in the image frame, or after converting the continuous image frame into a bird's-eye view, the effective range of the length and width as mentioned above is framed, that is, the first detection area 201 and the second detection area 202 are bird's-eye view areas projected on the ground.

[0042] (2) Step S022, determining whether the first detection area and the second detection area contain a recognition target. In the present invention, the processing unit 107 has an artificial intelligence for image recognition, and thus can determine whether the detection area (the first detection area 201 and the second detection area 202) contains a recognition target. The recognition target can be set as a single setting or multiple conditions, and can be set in an intersection or union manner according to the multiple conditions, that is, two or more different recognition targets can be set. The intersection represents the simultaneous existence of different recognition targets, and the union represents the existence of any recognition target. The system can be determined by the developer. In general, the recognition targets can include pedestrians, various vehicles, ground signs, lane lines, sidewalks, lane types, curbstones, parking spaces, ground locks, limit bars, etc. In the present invention, since a parking lot map is established, a simple and accurate recognition target is required, and thus most of them are fixed objects in the parking lot, such as parking spaces with grid lines, parking spaces without grid lines, parking space numbers, ground locks, limit bars, ground signs, and parking lot columns, curve lines, etc.

[0043] (3) Step S023, generating two-dimensional feature points or three-dimensional feature points of continuous image frames. When the result of the recognition by the processing unit 107 is: (i) when the first detection area 201 and the second detection area 202 include the recognition target in step S022, the processing unit 107 generates two-dimensional feature points with depth information for each frame of continuous image information captured by the camera 103 on the left side of the vehicle and the camera 104 on the right side. Such two-dimensional feature points with depth information are calculated based on the distance relative to the camera based on the internal and external parameters of the camera itself. The two-dimensional feature points are usually the corner points of the parking space, the sampling points of the parking space grid lines, the edges of the pillars or the edge sampling points of the parking space numbers, etc. It is worth noting that in this case In this case, the system 100 does not extract the three-dimensional feature points of the continuous image information of the vehicle's surrounding environment in step S01 to reduce the amount of calculation of the processing unit 107; (ii) when the processing unit 107 identifies that the recognition target in step S022 is not included in the first detection area 201 or the second detection area 202, the processing unit 107 will be able to cover all the continuous image information of the vehicle's surrounding environment, that is, the three-dimensional feature points with depth information are generated in each frame of the continuous image information of the vehicle's surrounding environment taken by the front camera 101, the rear camera 102, the left camera 103 and the right camera 104, such as Figure 5 When the vehicle moves from position A to position B, that is, within the first detection area 201 or the second detection area 202, if one of them fails to detect the identified target (position B), or the two-dimensional feature points fail to reach the threshold that satisfies the establishment of the two-dimensional point cloud map, the processing unit 107 generates three-dimensional feature points with depth information in each frame of all continuous image information of the vehicle's surroundings.

[0044] In detail, the processing unit 107 extracts three-dimensional feature points from continuous image frames in a model-based manner instead of computer vision, and mainly estimates the descriptors of each feature point in each frame of the image, especially the points, corners, and textures of common pillars, walls, road signs, symbols, etc. (but not limited to the above) in the parking lot area. However, it is worth noting that in order to effectively establish a parking lot map, especially the loop connection of autonomous parking mapping, the system 100 of the present invention starts the mapping mode and the vehicle drives within a preset starting distance for mapping (such as a distance of 20 meters, or set by the system developer). The processing unit 107 will simultaneously generate two-dimensional and three-dimensional feature points with depth information for each frame of continuous image information of the vehicle's surrounding environment.

[0045] Step S03, matching feature points with vehicle driving data information and constructing a point cloud map. Since the image information is a record composed of continuous frames, the front and rear frames of the image reflect the time difference. In step S02, even if the processing unit 107 generates two-dimensional or three-dimensional feature points with depth information for the continuous image information, there will be a certain difference between the actual vehicle driving distance, and as the vehicle driving distance increases, the cumulative error of the image itself will occur. Therefore, in this embodiment, the processing unit 107 receives the driving data information from the vehicle through the vehicle information assembly interface 105 according to step S01, especially the odometer information (usually composed of inertial measurement unit (Inertial Measurement Unit), wheel speed meter or RTK information, etc., which is related to the sensing information that can measure the vehicle moving distance). When the vehicle moving distance corresponding to the front and rear frames of the image is known, the two-dimensional or three-dimensional feature points can be matched with the vehicle driving information according to the triangulation principle to obtain the depth information of the actual feature points. In addition, in this step, the processing unit 107 will match the vehicle driving trajectory generated in the aforementioned step S01 and the feature points with depth information generated in each frame of the continuous image information of the vehicle's surrounding environment generated in step S02, so that these two-dimensional and three-dimensional feature points can have relative coordinates relative to the moving vehicle, and a two-dimensional point cloud map or / and a three-dimensional point cloud map can be constructed. In addition, for the three-dimensional feature points, the processing unit 107 in the present invention has semantic operations, which can remove dynamic feature points in the three-dimensional point cloud map. The purpose is that for the user's first map construction, the parking lot area may include many dynamic objects, such as parked vehicles, pedestrians, movable fire-fighting equipment, etc. If these dynamic objects are regarded as feature points and become part of the three-dimensional point cloud map during map construction, when the user subsequently turns on the autonomous parking function, the position of the dynamic objects changes or moves, causing the positioning of the autonomous parking function to fail. Therefore, through the semantic operation of the processing unit 107, the dynamic feature points in the area framed by the semantic operation can be deleted in each frame of the continuous image information of the vehicle's surrounding environment, that is, the information with dynamic feature points is not stored in the three-dimensional point cloud map. When the user performs the autonomous parking function next time, the system can again add these deleted dynamic feature points to the area of ​​the three-dimensional point cloud map, and appropriately update these areas that originally did not have feature points in the three-dimensional point cloud map based on the new feature points generated by the image information captured at that time, thereby improving the accuracy of the parking area map. In order to establish a better three-dimensional point cloud map, in addition to establishing a three-dimensional point cloud map based on the corresponding continuous image frames when the processing unit 107 fails to identify the target in the first detection area 201 or the second detection area 202, feature points with depth information are generated at the same time within a predetermined time before or after these corresponding continuous image frames (and several frames before and after) to construct a three-dimensional point cloud map, that is Figure 6Within a predetermined time (or several frames before and after) before or after the vehicle travels to position B, feature points with depth information can be obtained for these continuous image frames. Therefore, according to the present invention, a three-dimensional point cloud map can be constructed by continuous image frames that do not contain the identified target and continuous image frames that contain the identified target obtained within a predetermined time before and / or after determining that the identified target is not included in the first detection area 201 or the second detection area 202, so that the several continuous frames before and after include both the two-dimensional point cloud map and the three-dimensional point cloud map. Similarly, such a predetermined time (or number of frames) can be determined by the system developer.

[0046] Therefore, after this step, when the user activates the mapping function, the processing unit 107 can store these two-dimensional point cloud maps or / and three-dimensional point cloud maps in the storage module 108 by computing the images of the surrounding environment of the vehicle captured by the camera during the driving process of the vehicle. The stored information can include the continuous image information captured by each monocular camera, as well as the relative coordinates of these monocular cameras corresponding to the two-dimensional point cloud map and the three-dimensional point cloud map, the vehicle odometer information and the driving trajectory of the vehicle. The storage module 108 can even store the maps created by the user in different parking lots, or the maps created at different times in the same parking lot. Furthermore, the mapping completeness index of each parking lot after semantic computing of dynamic feature points, as well as the GPS information from the vehicle, can be stored. The information stored in the storage module 108 can be read, rewritten or transmitted to the cloud server through the network for the purpose of optimizing maps and information exchange.

[0047] Step S04, when it is determined that the driving trajectory of the vehicle is within an overlapping range, the stored point cloud map is matched and then spliced ​​to generate a corrected point cloud map. After the mapping function is started, since the processing unit 107 in step S01 receives driving data information from the vehicle during the driving of the vehicle, especially the odometer information, to generate the driving trajectory of the vehicle and generate the driving coordinates of the vehicle in the parking lot. Furthermore, the processing unit 107 can use RTK information to generate the driving trajectory. As the vehicle travels, in addition to the two-dimensional point cloud map and the three-dimensional point cloud map being stored, the driving data information is also continuously received by the processing unit 107 for calculation and synchronously stored in the storage module 108. Therefore, the coordinates of the vehicle when it moves can be updated through the processing unit 107's judgment of the odometer information, or the processing unit 107 determines the driving trajectory of the vehicle by comparing an RTK information. When the processing unit 107 determines that the vehicle's driving trajectory is within an overlapping range, it executes the extraction of two-dimensional and three-dimensional feature points for each frame of continuous image information of the vehicle's surroundings, and matches the stored point cloud map and performs splicing to generate a corrected point cloud map. Among them, in the present invention, the overlapping range can correspond to the effective field of view of the continuous image frames of the vehicle's surroundings that can be spliced ​​into a full field of view by the camera, and the overlapping range can be a distance range within 20 meters of the monocular camera field of view used in the present invention. In detail, if the coordinates of the vehicle during driving are within a distance of 20 meters from the vehicle's driving trajectory in step S01, the processing unit 107 will calculate and extract the two-dimensional and / or three-dimensional feature points of each frame of the continuous image information of all vehicle surroundings at this time. Similarly, the method of extracting feature points is the same as the aforementioned step S02. Please refer to Figure 7 , because the driving data information (odometer information) is constantly used to compare the stored driving trajectory after the mapping program is started, Figure 7 When the vehicle is circling the parking lot, as long as the driving trajectory is within the overlapping range, the processing unit 107 needs to extract the two-dimensional or / and three-dimensional feature points of each frame of the continuous image information of the surrounding environment of all vehicles at this time to match the stored two-dimensional or three-dimensional point cloud map. It is worth noting that since the point cloud map of the present invention is a panoramic point cloud map generated by the continuous image information that can cover all the surrounding environments of the vehicle, when the vehicle posture changes (such as Figure 7(a vehicle passes through an intersection or a starting point, and the vehicle is traveling in a different direction). Through the panoramic point cloud map of the present invention, a two-dimensional or / and three-dimensional feature point matching can be performed, and the matching result will not be affected by the vehicle posture. Once the processing unit 107 calculates that the matching degree of the feature points reaches the system setting value, it will perform operations such as splicing the vehicle's driving trajectory and the feature points in the parking lot, or delete unnecessary feature points. This can reduce the duplication or mismatch generated during map construction, achieve better loop connection, and finally generate a corrected point cloud map. After the above steps are completed, the Gauss-Newton method can be used to correct the point cloud map generated in the above steps for the optimization of the local or global map, which will not be repeated in the present invention.

Claims

1. A mapping correction system, installed on a vehicle, comprising: An image information receiving module receives continuous image information covering all surroundings of the vehicle captured by at least four monocular cameras installed in front of, behind, and on both sides of the vehicle; A processing unit receives driving data information to generate a driving trajectory of the vehicle; and when the vehicle performs a mapping procedure, the processing unit generates feature points with depth information in each frame of continuous image information of the vehicle's surrounding environment, and after matching the feature points with the driving data information of the vehicle, constructs and stores a point cloud map; It is characterized in that when the processing unit determines that the driving trajectory of the vehicle is within an overlapping range, the stored point cloud map is matched and then spliced ​​to generate a corrected point cloud map.

2. The map correction system according to claim 1, characterized in that: The point cloud map includes a two-dimensional point cloud map and / or a three-dimensional point cloud map.

3. The map correction system according to claim 1, characterized in that: The driving data information is odometer information.

4. The map correction system according to claim 2, characterized in that: The processing unit is used to define a detection area of ​​continuous image information of the vehicle's surrounding environment; determine whether the detection area contains an identification target, wherein when the detection area includes the identification target, the processing unit constructs the two-dimensional point cloud map corresponding to the identification target; when the identification target is not included in the detection area, the processing unit constructs the three-dimensional point cloud map.

5. The map correction system according to claim 4, characterized in that: The detection area is a bird's-eye view area projected on the ground.

6. The map correction system according to claim 1, characterized in that: The processing unit determines whether the driving track of the vehicle is within an overlapping range by comparing a real-time dynamic positioning information.

7. The map correction system according to claim 2, characterized in that: The processing unit removes dynamic feature points in the three-dimensional point cloud map through semantic operations.