Parking method and device based on three-dimensional parking space

Abstract

The present application discloses a parking method and device based on a three-dimensional parking space. After acquiring an image around the vehicle, since a three-dimensional parking space generally does not have parking space line information, the vacant three-dimensional parking space is no longer identified in the image based on the parking space line information. Instead, the vacant three-dimensional parking space is identified in the image based on two corner points on the parking entry side of the three-dimensional parking space. If the corner point distance between the two corner points meets a first threshold condition, the vehicle can be parked in the vacant three-dimensional parking space, and the vacant three-dimensional parking space is determined as the target parking space. In order to correctly park the vehicle in the target parking space, valid point cloud data representing the position of the guard edges on both sides of the target parking space is obtained based on a Tof camera. Based on the valid point cloud data, the relative position relationship between the target parking space and the vehicle can be determined, so that the vehicle can be controlled to park in the target parking space based on a point cloud function, thereby improving the detection efficiency of the target parking space and the accuracy of parking.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a parking method and device based on a three-dimensional parking space. Background Art

[0002] With the development of science and technology and the improvement of people's living standards, the use and purchase of cars have greatly increased. However, with the sharp increase in the number of cars, the difficulty of parking due to the large number of vehicles and the lack of parking spaces has become a problem in urban transportation. In order to make full use of urban space and increase parking spaces as much as possible, more and more parking lots are using multi-story parking spaces.

[0003] To assist users in parking, vehicles currently generally use ultrasonic radar detection, that is, ultrasonic radar detects obstacle information in a specific area to detect parking spaces. Under ideal conditions, the ultrasonic radar will return an empty parking space between two vehicles, and then screen and determine whether the empty parking space can be parked based on the returned information, thereby realizing automatic parking.

[0004] However, since ultrasonic radar detection of empty parking spaces depends on the presence of vehicles or large objects on both sides of the empty parking spaces, and the special shape of multi-story parking spaces, with large intervals and no vehicles or large objects on both sides, the ultrasonic radar detection method cannot detect multi-story parking spaces, and thus cannot achieve the purpose of assisting users in parking. Summary of the Invention

[0005] In response to the above problems, the present application provides a parking method and device based on a three-dimensional parking space, which are used to detect three-dimensional parking spaces and assist users in parking.

[0006] A first aspect of an embodiment of the present application provides a parking method based on a three-dimensional parking space, the method comprising:

[0007] Acquire images around the vehicle;

[0008] Identify the corner point distance of an empty three-dimensional parking space based on the image, where the corner point distance is the distance between two corner points on the parking entrance side of the three-dimensional parking space;

[0009] Determine an empty parking space whose corner point distance satisfies a first threshold condition as a target parking space, wherein the first threshold condition is determined according to the width of the vehicle;

[0010] Based on the time-of-flight (Tof) camera, valid point cloud data representing the positions of the guard edges on both sides of the target parking space is obtained;

[0011] Determine a point cloud function corresponding to the target parking space according to the valid point cloud data, wherein the point cloud function is a functional representation of the center line of the target parking space;

[0012] The vehicle is controlled to park in the target parking space based on the point cloud function.

[0013] Optionally, the method further includes:

[0014] Obtaining wheel pulses and steering wheel angles of the vehicle;

[0015] determining a track of the vehicle based on the wheel pulses and the steering wheel angle;

[0016] determining a deviation between the track and the point cloud function;

[0017] The controlling the vehicle to park in the target parking space based on the point cloud function includes:

[0018] If the deviation satisfies a second threshold condition, the vehicle is controlled to park in the target parking space based on the point cloud function.

[0019] If the position does not satisfy the second threshold condition, adjust the position of the vehicle until the deviation between the adjusted track of the vehicle and the point cloud function satisfies the second threshold condition.

[0020] Optionally, obtaining valid point cloud data of the guard edge positions on both sides of the target parking space based on the time-of-flight (Tof) camera includes:

[0021] Obtaining point cloud data representing the target parking space through the Tof camera;

[0022] The point cloud data are clustered based on the height information to obtain valid point cloud data of the guard edge positions on both sides of the target parking space.

[0023] Optionally, acquiring images around the vehicle includes:

[0024] A four-way surround-view fisheye stitched image representing the vehicle's surrounding environment is obtained based on a surround-view fisheye camera;

[0025] The step of identifying the corner point distance of an idle three-dimensional parking space according to the image includes:

[0026] The corner point distance between two corner points on the parking entry side of the vacant three-dimensional parking space is determined according to the four-way surround view fisheye stitching image.

[0027] Optionally, the first threshold condition is determined based on the width of the vehicle and the imbalance ratio of the surround-view fisheye camera.

[0028] Optionally, determining the vacant multi-story parking space whose corner point distance satisfies a first threshold condition as the target parking space includes:

[0029] Get the corner point distances corresponding to m vacant parking spaces;

[0030] Determine n vacant three-dimensional parking spaces whose distances from the m corner points meet a first threshold condition as pending three-dimensional parking spaces;

[0031] The distances between the undetermined three-dimensional parking spaces and the vehicle are determined, and a three-dimensional parking space whose distance satisfies a third threshold condition is determined as a target parking space.

[0032] A second aspect of an embodiment of the present application provides a parking device based on a three-dimensional parking space, the device comprising: an acquisition unit, an identification unit, a target parking space determination unit, a point cloud data acquisition unit, a point cloud function determination unit, and a control unit;

[0033] The acquisition unit is used to acquire images around the vehicle;

[0034] The recognition unit is configured to recognize a corner point distance of an empty three-dimensional parking space based on the image, wherein the corner point distance is a distance between two corner points on a parking entry side of the three-dimensional parking space;

[0035] The target parking space determining unit is configured to determine an empty three-dimensional parking space whose corner point distance satisfies a first threshold condition as a target parking space, wherein the first threshold condition is determined according to the width of the vehicle;

[0036] The point cloud data acquisition unit is used to obtain valid point cloud data representing the positions of the guard edges on both sides of the target parking space based on a time-of-flight (Tof) camera;

[0037] The point cloud function determining unit is configured to determine a point cloud function corresponding to the target parking space based on the valid point cloud data, wherein the point cloud function is a functional representation of a center line of the target parking space;

[0038] The control unit is configured to control the vehicle to park in the target parking space based on the point cloud function.

[0039] Optionally, the device further includes a deviation determining unit, configured to:

[0040] Obtaining wheel pulses and steering wheel angles of the vehicle;

[0041] determining a track of the vehicle based on the wheel pulses and the steering wheel angle;

[0042] determining a deviation between the track and the point cloud function;

[0043] The control unit is used to:

[0044] If the deviation satisfies a second threshold condition, the vehicle is controlled to park in the target parking space based on the point cloud function.

[0045] If the position does not satisfy the second threshold condition, adjust the position of the vehicle until the deviation between the adjusted track of the vehicle and the point cloud function satisfies the second threshold condition.

[0046] Optionally, the point cloud data acquisition unit is used to:

[0047] Obtaining point cloud data representing the target parking space through the Tof camera;

[0048] The point cloud data are clustered based on the height information to obtain valid point cloud data of the guard edge positions on both sides of the target parking space.

[0049] Optionally, the acquiring unit is configured to:

[0050] Acquire four-way surround-view fisheye stitched images representing the vehicle's surrounding environment based on the surround-view fisheye camera;

[0051] The identification unit is used to:

[0052] A corner point distance between two corner points on the parking entry side of the vacant three-dimensional parking space is determined according to the four-way surround view fisheye stitched image.

[0053] Compared with the prior art, the advantages of the above technical solution of this application are:

[0054] After acquiring an image of the vehicle's surroundings, since 3D parking spaces generally lack parking space line information, vacant 3D parking spaces are no longer identified in the image based on parking space line information. Instead, vacant 3D parking spaces are identified based on two corner points on the parking entrance side of the 3D parking space. If the distance between the two corner points meets a first threshold, the vehicle can park in the vacant 3D parking space and the vacant 3D parking space is identified as the target parking space. To accurately park the vehicle in the target parking space, valid point cloud data representing the position of the guardrails on both sides of the target parking space is acquired using a Tof camera. Based on this valid point cloud data, the relative positional relationship between the target parking space and the vehicle can be determined, allowing the vehicle to be controlled to park in the target parking space based on a point cloud function. Thus, identifying vacant 3D parking spaces based on the two corner points on the parking entrance side of the 3D parking space quickly narrows the range and improves the detection efficiency of the target parking space. During parking, the Tof camera's secondary sensing allows for precise positioning, independent of the presence of vehicles or larger objects on either side of the target parking space, thus improving parking accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 A flowchart of a parking method based on a three-dimensional parking space provided in an embodiment of the present application;

[0057] Figure 2 A schematic diagram of a parking device based on a three-dimensional parking space provided in this application. DETAILED DESCRIPTION

[0058] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0059] In order to assist users in parking, relevant technologies not only use ultrasonic radar detection but also pure visual detection. After acquiring images around the vehicle, the four corner points and parking line information of the parking space are identified in the image to determine the parking space.

[0060] However, purely visual inspection relies on strong color or contrast characteristics along the four edges of the parking space. Currently, there are no comprehensive laws and regulations regarding multi-story parking spaces. Many parking lots, to save costs, do not draw parking space lines on multi-story parking spaces. Furthermore, since most multi-story parking spaces are made of metal, it is difficult to ensure strong color and contrast along the edges, resulting in low detection accuracy for purely visual inspection. Furthermore, the accuracy of purely visual inspection relies on the ground being flat and the parking space being at the same height as the ground. Multi-story parking spaces often have a slope, and the parking plane of multi-story parking spaces is at a certain height from the ground. This results in low detection accuracy for purely visual inspection, making it impossible to achieve the high-precision detection and positioning requirements of narrow multi-story parking spaces.

[0061] Based on this, embodiments of the present application provide a parking method based on a three-dimensional parking space. After acquiring an image of the vehicle's surroundings, since three-dimensional parking spaces generally lack parking space line information, vacant three-dimensional parking spaces are no longer identified based on parking space line information. Instead, the method identifies vacant three-dimensional parking spaces based on the two corner points on the parking entrance side of the three-dimensional parking space. If the corner point distance between the two corner points meets a first threshold condition, the vehicle can park in the vacant three-dimensional parking space and the vacant three-dimensional parking space is determined as the target parking space. To ensure accurate parking of the vehicle in the target parking space, valid point cloud data representing the position of the guard rails on both sides of the target parking space is acquired using a Tof camera. Based on this valid point cloud data, the relative positional relationship between the target parking space and the vehicle can be determined, thereby controlling the vehicle to park in the target parking space based on a point cloud function. Thus, by identifying vacant three-dimensional parking spaces based on the two corner points on the parking entrance side of the three-dimensional parking space, the range can be quickly narrowed, improving the detection efficiency of the target parking space. During parking, the Tof camera provides secondary sensing and precise positioning, independent of the presence of vehicles or larger objects on both sides of the target parking space, thereby improving parking accuracy.

[0062] The parking method based on a three-dimensional parking space provided in an embodiment of the present application is described in detail below with reference to the accompanying drawings.

[0063] See also Figure 1 , Figure 1 This is a flow chart of a parking method based on a three-dimensional parking space provided by the present application. The method may include the following steps 101-106.

[0064] S101: Acquire images around the vehicle.

[0065] In order to detect an empty three-dimensional parking space, an image around the vehicle may be acquired first.

[0066] As a possible implementation method, when the vehicle is a certain distance away from the multi-story parking space, such as 5-10 meters longitudinally and 1-2 meters laterally, and the vehicle's forward direction is perpendicular to the parking direction of the multi-story parking space, images around the vehicle are obtained to detect vacant multi-story parking spaces.

[0067] As a possible implementation method, a four-way surround-view fisheye stitched image that can represent the vehicle's surrounding environment can be obtained based on a surround-view fisheye camera. The surround-view fisheye camera has a large viewing angle of more than 180° and has good perception at close distances. It can be installed under the left and right rearview mirrors and under the front and rear license plates of the vehicle to perform image stitching, parking space detection, visualization and other functions.

[0068] S102: Identify the corner point distances of the vacant three-dimensional parking spaces based on the image.

[0069] After obtaining an image of the vehicle's surroundings, an empty three-dimensional parking space may be identified based on the image, thereby determining a corner point distance between two corner points on the parking entry side of the empty three-dimensional parking space.

[0070] As a possible implementation method, an image, such as a four-way surround view fisheye stitched image, can be input into a deep learning model, and the information of the two corner points can be obtained through the deep learning model to determine the corner point distance.

[0071] Among them, Deep Learning (DL) is a new research direction in the field of Machine Learning (ML). It is introduced into machine learning to make it closer to its original goal - Artificial Intelligence (AI).

[0072] S103: Determine an empty multi-story parking space whose corner point distance meets a first threshold condition as a target parking space.

[0073] After obtaining the corner point distance, it is necessary to determine whether the vehicle can be parked in the three-dimensional parking space corresponding to the corner point distance. Therefore, it is determined whether the corner point distance meets the first threshold condition. If it does, the space of the three-dimensional parking space is large enough to park the vehicle, and it is determined as the target parking space. If it does not meet the condition, the space of the three-dimensional parking space is too small to park the vehicle, thereby determining whether the vehicle can be parked in the three-dimensional parking space based on the corner point distance.

[0074] The first threshold condition is determined according to the width of the vehicle, so as to ensure that the width of the vehicle does not exceed the width of the multi-story parking space, so that the vehicle can be parked correctly in the multi-story parking space.

[0075] Therefore, using the simplest corner point information to determine the target parking space can overcome the problem of pure visual detection in related technologies that has detection failures due to strong light, shadows, and partial occlusion. It can also provide prior information for the position of the subsequent vehicle's rear end to align with the parking side of the three-dimensional parking space. In addition, the Tof camera used subsequently has a very small field of view, which can quickly determine the target parking space, improve the detection rate of single-frame images and the robustness of multiple scenarios.

[0076] As a possible implementation, if the image of the vehicle's surroundings is obtained using a surround-view fisheye camera, the threshold can be determined based on the camera's imbalance ratio and the vehicle's width. Because fisheye cameras are wide-angle cameras, they exhibit certain distortion. Furthermore, as previously mentioned, the parking plane of a stereoscopic parking space is at a certain height above the ground. This causes the corner distances determined based on the image to be smaller than the actual corner distances. Therefore, the first threshold condition can be determined based on the camera's imbalance ratio and the vehicle's width. For example, the first threshold condition can be set between a lower limit distance value and an upper limit distance value, with the lower limit distance value being lower than the vehicle width and the upper limit distance value being higher. This improves the accuracy of stereoscopic parking space detection.

[0077] For example, if a vehicle is 2 meters wide, the first threshold condition can be set between 1.8 and 2.2 meters. Even if the corner distance determined based on the image is 1.9 meters, which is smaller than the vehicle's width, it can still be identified as the target parking space. This first threshold condition overcomes the error caused by the surround-view fisheye camera's distortion and improves the utilization rate of vacant parking spaces.

[0078] As one possible implementation, if the corner point distances corresponding to m available 3D parking spaces are obtained based on an image, n available 3D parking spaces can be screened from the m available 3D parking spaces based on a first threshold condition and identified as pending 3D parking spaces, where n is less than or equal to m. Based on the distance between the pending 3D parking spaces and the vehicle, the 3D parking spaces whose distances meet a third threshold condition are identified as target parking spaces. For example, the n pending 3D parking spaces can be sorted by distance, and the 3D parking space with the smallest distance, i.e., the closest to the vehicle, can be identified as the target parking space. Alternatively, the n pending 3D parking spaces can be displayed on an onboard computer, and the target parking space can be determined based on the user's selection.

[0079] S104: Based on the time-of-flight (Tof) camera, valid point cloud data representing the positions of the guard edges on both sides of the target parking space is obtained.

[0080] It should be noted that the vehicle control module performs path planning based on the two corner points of the multi-story parking space determined above and the midpoint formed by the two corner points, aiming to align the rear of the vehicle in front of the parking entrance of the multi-story parking space so that subsequent vehicles can park in the target parking space.

[0081] In order to park the vehicle correctly in the target parking space, the location of the target parking space needs to be determined. Since the protective edges on both sides of the three-dimensional parking space are relatively high, the effective point cloud data corresponding to the position of the protective edges on both sides of the target parking space can be obtained based on the Time of Flight (Tof) camera. Among them, the Tof camera emits modulated near-infrared light through the sensor. After encountering an object, it is reflected. The sensor calculates the time difference or phase difference between the emission and reflection of the light to convert the distance of the photographed scene to generate depth information. In addition, combined with traditional camera shooting, the three-dimensional outline of the object can be presented in the form of a topographic map with different colors representing different distances.

[0082] As one possible implementation, the target parking space includes at least the guardrails on both sides of the three-dimensional parking space and the parking plane of the three-dimensional parking space. Point cloud data representing the target parking space is acquired using a Tof camera. This point cloud data includes point cloud data corresponding to the guardrails and the parking plane of the three-dimensional parking space. Because the guardrails and the parking plane of the three-dimensional parking space are at different heights, the acquired point cloud data can be clustered based on this height information to filter out valid point cloud data representing the guardrails and achieve noise reduction.

[0083] S105: Determine a point cloud function corresponding to the target parking space based on the valid point cloud data.

[0084] Among them, the point cloud function is the function representation of the center line of the target parking space.

[0085] As a possible implementation method, the least squares method can be used to fit the effective point cloud data to determine the point cloud function.

[0086] S106: Controlling the vehicle to park in the target parking space based on the point cloud function.

[0087] Therefore, the combination of a Tof camera requires only corner information for the initial detection of a 3D parking space, significantly improving the success rate of 3D parking space detection. Furthermore, using the Tof camera for secondary positioning significantly improves the accuracy of 3D parking spaces with certain height differences and slopes. Furthermore, the system can park in confined spaces without relying on distinct color or contrast characteristics of the 3D parking space, or the presence of large parking spaces on either side of the vacant space.

[0088] As a possible implementation method, when the vehicle control module aligns the rear end of the vehicle in front of the entrance to the multi-story parking space, the vehicle will adjust its position, which will cause the vehicle to be located far away from the target parking space, thereby losing real-time detection of the target parking space. Therefore, dead reckoning is introduced. Among them, dead reckoning is a method of obtaining the track and information about the vehicle's surroundings based on the vehicle's turning angle and wheel speed without the help of external navigation objects. It should be noted that the device for real-time detection of the target parking space here can be a surround-view fisheye camera. Due to the possibility of distortion, the corner points captured by the surround-view fisheye camera may be distorted, etc., resulting in the inability to capture the corner points, and then the problem of not being able to detect the target parking space.

[0089] Therefore, before S106, the vehicle's wheel pulses and steering wheel angles can be obtained, and the vehicle's track can be determined based on the wheel pulses and steering wheel angles. For example, the vehicle's track can be predicted by dead reckoning using a vehicle motion model, integrating the vehicle's forward distance and angle. The deviation between the track and the point cloud function is then determined. If the deviation satisfies a second threshold condition, it indicates that the vehicle's current position allows parking in the target parking space, and the vehicle is controlled to park in the target parking space based on the point cloud function. If the deviation does not meet the second threshold condition, it indicates that the vehicle's current position does not allow parking in the target parking space, so the vehicle's position is adjusted, and the vehicle's wheel pulses and steering wheel angles are again obtained to determine the adjusted track of the vehicle. The deviation between the two is then determined until the deviation between the adjusted track and the point cloud function satisfies the second threshold condition.

[0090] The second threshold condition is the allowable deviation, which can be a deviation of 5 cm between the track and the point cloud function. For example, if a perpendicular line drawn from the track's center to the point cloud function is no longer than 5 cm, the vehicle is considered to be able to park in the target space.

[0091] After acquiring an image of the vehicle's surroundings, since 3D parking spaces generally lack parking space line information, vacant 3D parking spaces are no longer identified in the image based on parking space line information. Instead, vacant 3D parking spaces are identified based on two corner points on the parking entrance side of the 3D parking space. If the distance between the two corner points meets a first threshold, the vehicle can park in the vacant 3D parking space and the vacant 3D parking space is identified as the target parking space. To accurately park the vehicle in the target parking space, valid point cloud data representing the position of the guardrails on both sides of the target parking space is acquired using a Tof camera. Based on this valid point cloud data, the relative positional relationship between the target parking space and the vehicle can be determined, allowing the vehicle to be controlled to park in the target parking space based on a point cloud function. Thus, identifying vacant 3D parking spaces based on the two corner points on the parking entrance side of the 3D parking space quickly narrows the range and improves the detection efficiency of the target parking space. During parking, the Tof camera's secondary sensing allows for precise positioning, independent of the presence of vehicles or larger objects on either side of the target parking space, thus improving parking accuracy.

[0092] In addition to providing a parking method based on a three-dimensional parking space, the embodiment of the present application also provides a parking device based on a three-dimensional parking space, such as Figure 2 As shown, it includes: an acquisition unit 201, an identification unit 202, a target parking space determination unit 203, a point cloud data acquisition unit 204, a point cloud function determination unit 205 and a control unit 206;

[0093] The acquisition unit 201 is used to acquire images around the vehicle;

[0094] The recognition unit 202 is configured to recognize a corner point distance of an empty three-dimensional parking space based on the image, wherein the corner point distance is a distance between two corner points on the parking entry side of the three-dimensional parking space;

[0095] The target parking space determining unit 203 is configured to determine an empty parking space whose corner point distance satisfies a first threshold condition as a target parking space, where the first threshold condition is determined according to the width of the vehicle;

[0096] The point cloud data acquisition unit 204 is configured to obtain valid point cloud data representing the positions of the guard edges on both sides of the target parking space based on a time-of-flight (Tof) camera;

[0097] The point cloud function determining unit 205 is configured to determine a point cloud function corresponding to the target parking space based on the valid point cloud data, wherein the point cloud function is a functional representation of the center line of the target parking space;

[0098] The control unit 206 is configured to control the vehicle to park in the target parking space based on the point cloud function.

[0099] As a possible implementation, the apparatus further includes a deviation determining unit 207, configured to:

[0100] Obtaining wheel pulses and steering wheel angles of the vehicle;

[0101] determining a track of the vehicle based on the wheel pulses and the steering wheel angle;

[0102] determining a deviation between the track and the point cloud function;

[0103] The control unit 206 is configured to:

[0104] If the deviation satisfies a second threshold condition, the vehicle is controlled to park in the target parking space based on the point cloud function.

[0105] If the position does not satisfy the second threshold condition, adjust the position of the vehicle until the deviation between the adjusted track of the vehicle and the point cloud function satisfies the second threshold condition.

[0106] As a possible implementation, the point cloud data acquisition unit 204 is configured to:

[0107] Obtaining point cloud data representing the target parking space through the Tof camera;

[0108] The point cloud data are clustered based on the height information to obtain valid point cloud data of the guard edge positions on both sides of the target parking space.

[0109] As a possible implementation, the acquiring unit 201 is configured to:

[0110] Acquire four-way surround-view fisheye stitched images representing the vehicle's surrounding environment based on the surround-view fisheye camera;

[0111] The identification unit 202 is configured to:

[0112] A corner point distance between two corner points on the parking entry side of the vacant three-dimensional parking space is determined according to the four-way surround view fisheye stitched image.

[0113] As a possible implementation manner, the first threshold condition is determined according to the width of the vehicle and the imbalance ratio of the surround-view fisheye camera.

[0114] As a possible implementation, the target parking space determination unit 203 is configured to:

[0115] Get the corner point distances corresponding to m vacant parking spaces;

[0116] Determine n vacant three-dimensional parking spaces whose distances from the m corner points meet a first threshold condition as pending three-dimensional parking spaces;

[0117] The distances between the undetermined three-dimensional parking spaces and the vehicle are determined, and a three-dimensional parking space whose distance satisfies a third threshold condition is determined as a target parking space.

[0118] Thus, embodiments of the present application provide a parking device based on a three-dimensional parking space. After acquiring an image of the vehicle's surroundings, since three-dimensional parking spaces generally lack parking space line information, vacant three-dimensional parking spaces are no longer identified based on parking space line information. Instead, the device identifies vacant three-dimensional parking spaces based on the two corner points on the parking entrance side of the three-dimensional parking space. If the corner point distance between the two corner points meets a first threshold condition, the vehicle can park in the vacant three-dimensional parking space and the vacant three-dimensional parking space is determined as the target parking space. To accurately park the vehicle in the target parking space, a Tof camera is used to acquire valid point cloud data representing the position of the guard rails on both sides of the target parking space. Based on this valid point cloud data, the relative positional relationship between the target parking space and the vehicle can be determined, thereby controlling the vehicle's parking into the target parking space based on a point cloud function. Thus, by identifying vacant three-dimensional parking spaces based on the two corner points on the parking entrance side of the three-dimensional parking space, the range can be quickly narrowed, improving the detection efficiency of the target parking space. During parking, the Tof camera's secondary sensing accurately locates the vehicle, independent of the presence of vehicles or larger objects on both sides of the target parking space, thereby improving parking accuracy.

[0119] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely illustrative, and the units and modules described as separate components may or may not be physically separated. In addition, some or all of the units and modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0120] The above is only a specific implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A parking method based on a three-dimensional parking space, characterized in that: The method comprises: Acquire images around the vehicle; Identify the corner point distance of an empty three-dimensional parking space based on the image, where the corner point distance is the distance between two corner points on the parking entrance side of the three-dimensional parking space; Determine an empty parking space whose corner point distance satisfies a first threshold condition as a target parking space, wherein the first threshold condition is determined according to the width of the vehicle; Obtaining point cloud data representing the target parking space through a Tof camera; Clustering the point cloud data based on the height information to obtain valid point cloud data of the guard edge positions on both sides of the target parking space; Determine a point cloud function corresponding to the target parking space according to the valid point cloud data, wherein the point cloud function is a functional representation of the center line of the target parking space; Obtaining wheel pulses and steering wheel angles of the vehicle; determining a track of the vehicle based on the wheel pulses and the steering wheel angle; determining a deviation between the track and the point cloud function; If the deviation satisfies a second threshold condition, controlling the vehicle to park in the target parking space based on the point cloud function; If the position does not satisfy the second threshold condition, adjust the position of the vehicle until the deviation between the adjusted track of the vehicle and the point cloud function satisfies the second threshold condition.

2. The method according to claim 1, characterized in that The acquiring of images around the vehicle includes: Acquire four-way surround-view fisheye stitched images representing the vehicle's surrounding environment based on the surround-view fisheye camera; The step of identifying the corner point distance of an idle three-dimensional parking space according to the image includes: A corner point distance between two corner points on the parking entry side of the vacant three-dimensional parking space is determined according to the four-way surround view fisheye stitched image.

3. The method according to claim 2, characterized in that The first threshold condition is determined according to the width of the vehicle and the imbalance ratio of the surround-view fisheye camera.

4. The method according to claim 1, wherein The step of determining the vacant parking space whose corner point distance satisfies the first threshold condition as the target parking space includes: Get the corner point distances corresponding to m vacant parking spaces; Determine n vacant three-dimensional parking spaces whose distances from the m corner points meet a first threshold condition as pending three-dimensional parking spaces; The distances between the undetermined three-dimensional parking spaces and the vehicle are determined, and a three-dimensional parking space whose distance satisfies a third threshold condition is determined as a target parking space.

5. A parking device based on a three-dimensional parking space, characterized in that: The device comprises: an acquisition unit, an identification unit, a target parking space determination unit, a point cloud data acquisition unit, a point cloud function determination unit, a deviation determination unit and a control unit; The acquisition unit is used to acquire images around the vehicle; The recognition unit is configured to recognize a corner point distance of an empty three-dimensional parking space based on the image, wherein the corner point distance is a distance between two corner points on a parking entry side of the three-dimensional parking space; The target parking space determining unit is configured to determine an empty three-dimensional parking space whose corner point distance satisfies a first threshold condition as a target parking space, wherein the first threshold condition is determined according to the width of the vehicle; The point cloud data acquisition unit obtains point cloud data representing the target parking space through a Tof camera; Clustering the point cloud data based on the height information to obtain valid point cloud data of the guard edge positions on both sides of the target parking space; The point cloud function determining unit is configured to determine a point cloud function corresponding to the target parking space based on the valid point cloud data, wherein the point cloud function is a functional representation of a center line of the target parking space; The deviation determination unit is configured to obtain wheel pulses and steering wheel angles of the vehicle; determining a track of the vehicle based on the wheel pulses and the steering wheel angle; determining a deviation between the track and the point cloud function; the control unit being configured to control the vehicle to park in the target parking space based on the point cloud function if the deviation satisfies a second threshold condition; If the position does not satisfy the second threshold condition, adjust the position of the vehicle until the deviation between the adjusted track of the vehicle and the point cloud function satisfies the second threshold condition.

6. The device according to claim 5, characterized in that The acquisition unit is configured to: Acquire four-way surround-view fisheye stitched images representing the vehicle's surrounding environment based on the surround-view fisheye camera; The identification unit is used to: A corner point distance between two corner points on the parking entry side of the vacant three-dimensional parking space is determined according to the four-way surround view fisheye stitched image.

Images