Vehicle posture detection method, device and electronic equipment

Abstract

Embodiments of the present invention provide a vehicle posture detection method, device, and electronic device, relating to the field of robot visual perception technology. The method comprises: obtaining a body point cloud of a vehicle to be detected; wherein the body point cloud includes at least three wheels of the vehicle to be detected; extracting a bottom point cloud of the vehicle to be detected from the body point cloud; wherein the bottom point cloud is the point cloud between the vehicle chassis and the ground of the vehicle to be detected; determining the bottom edge point cloud of each wheel included in the body point cloud by identifying the wheels in the bottom point cloud; determining the landing point of each wheel based on the bottom edge point cloud of each wheel, and calculating the target posture of the vehicle to be detected based on the obtained landing points of each wheel. Compared with the prior art, the solution provided by the embodiment of the present invention achieves accurate acquisition of the vehicle's posture and wheel landing points, improves the accuracy of the parking robot's grip on the vehicle, and enables the parking robot to park the vehicle properly.

Description

Technical Field

[0001] The present invention relates to the field of robot visual perception technology, and in particular to a vehicle posture detection method, device and electronic equipment. Background Art

[0002] Currently, with the continuous development of robotics technology, parking robots are widely used in more and more parking lots, garages and other places.

[0003] A parking robot is a mobile robot used for automated parking. It can grip and carry a vehicle to an available parking space, replacing the driver in parking. To do this, the robot must first determine the vehicle's position.

[0004] In related art, the method for parking a vehicle by a parking robot typically involves installing a laser radar at a high point and a low point in the parking area. When a vehicle enters the parking area, the laser radar at the high point scans the vehicle to obtain its width information, while the laser radar at the low point measures the vehicle's wheel diameter and wheelbase information. The parking robot obtains the width information, wheel diameter, and wheelbase information from the two laser radars, and generates vehicle appearance information based on the width information, wheel diameter, and wheelbase information. In this way, the parking robot can clamp and carry the vehicle based on the generated vehicle appearance information to complete parking.

[0005] However, in the aforementioned related technologies, when the vehicle's direction of travel is at an angle to the laser radar's scanning area, the vehicle information obtained by the high-level laser radar and the low-level laser radar will have significant errors. This will seriously affect the parking robot's ability to accurately grasp the vehicle, resulting in the robot being unable to park the vehicle properly. Summary of the Invention

[0006] The purpose of the embodiments of the present invention is to provide a vehicle posture detection method, device, and electronic device to accurately obtain the vehicle posture and wheel contact points, improve the accuracy of the parking robot's grip on the vehicle, and enable the parking robot to park the vehicle normally. The specific technical solution is as follows:

[0007] In a first aspect, an embodiment of the present invention provides a vehicle posture detection method, the method comprising:

[0008] Obtaining a body point cloud of the vehicle to be detected; wherein the body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar;

[0009] Extracting a bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the bottom point cloud is: a point cloud between the vehicle chassis of the vehicle to be detected and the ground;

[0010] Determining the lower edge point cloud of each wheel included in the vehicle body point cloud by wheel recognition of the vehicle bottom point cloud;

[0011] Based on the lower edge point cloud of each wheel, the landing point of each wheel is determined, and based on the obtained landing points of each wheel, the target posture of the vehicle to be detected is calculated.

[0012] Optionally, in a specific implementation, obtaining a body point cloud of the vehicle to be detected includes:

[0013] Obtain scene point cloud of the vehicle to be detected;

[0014] Taking the ground as a reference plane, generating a depth map corresponding to the scene point cloud;

[0015] Performing a connected domain analysis on the depth map to obtain a maximum connected domain in the depth map;

[0016] A point cloud located in an area corresponding to the largest connected domain in the scene point cloud is extracted as a body point cloud of the vehicle to be detected.

[0017] Optionally, in a specific implementation, before extracting the point cloud located in the area corresponding to the maximum connected component from the scene point cloud, the method further includes:

[0018] Determining whether the size of the maximum connected domain falls within a preset vehicle size range;

[0019] If so, extract the point cloud in the scene point cloud that is located in the area corresponding to the largest connected domain.

[0020] Optionally, in a specific implementation, obtaining a scene point cloud related to the vehicle to be detected includes:

[0021] Obtaining an initial point cloud of a vehicle to be detected collected by at least one radar;

[0022] The obtained initial point clouds are spliced ​​to obtain the scene point cloud of the vehicle to be detected.

[0023] Optionally, in a specific implementation, extracting the bottom point cloud of the vehicle to be detected from the vehicle body point cloud includes:

[0024] Extracting a point cloud located in a specified area from the vehicle body point cloud as a calibration point cloud;

[0025] Determining a chassis height of the vehicle to be detected based on a specified ratio of the calibration point cloud and the specified area;

[0026] From the vehicle body point cloud, the point cloud below the chassis height is extracted as the bottom point cloud of the vehicle to be detected.

[0027] Optionally, in a specific implementation, the calculating the chassis height of the vehicle to be detected based on the calibration point cloud and the specified ratio matching the specified area includes:

[0028] Generating a height histogram of the calibration point cloud;

[0029] Based on the height histogram, determining each height interval having a calibration ratio equal to a specified ratio matching the specified area; wherein the calibration ratio of each height interval is: a ratio of the number of points included in the height interval to the number of points included in the specified area;

[0030] The upper limit value of the minimum height interval in each height interval is determined as the chassis height of the vehicle to be detected.

[0031] Optionally, in a specific implementation, determining the lower edge point cloud of each wheel included in the vehicle body point cloud by identifying the wheels of the vehicle bottom point cloud includes:

[0032] Based on a preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud, clustering the vehicle bottom point cloud to obtain the position of each wheel;

[0033] According to the position of each wheel, the lower edge point cloud of each wheel is extracted from the vehicle bottom point cloud.

[0034] Optionally, in a specific implementation, determining the landing point of each wheel based on the lower edge point cloud of each wheel includes:

[0035] Determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel; or,

[0036] Performing circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determining the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

[0037] Optionally, in a specific implementation, the method further includes:

[0038] The landing point of each wheel and the target posture are sent to a preset parking robot, so that the parking robot uses the landing point of each wheel as the optimal clamping point and parks the vehicle to be detected based on the target posture.

[0039] In a second aspect, an embodiment of the present invention provides a vehicle posture detection device, the device comprising:

[0040] A vehicle body point cloud acquisition module, configured to acquire a vehicle body point cloud of the vehicle to be detected; wherein the vehicle body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar;

[0041] A vehicle bottom point cloud extraction module is used to extract the vehicle bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the vehicle bottom point cloud is: the point cloud between the vehicle chassis of the vehicle to be detected and the ground;

[0042] a lower edge point cloud extraction module, configured to determine the lower edge point cloud of each wheel included in the vehicle body point cloud by identifying the wheels of the vehicle bottom point cloud;

[0043] The vehicle posture calculation module is used to determine the landing point of each wheel based on the lower edge point cloud of each wheel, and calculate the target posture of the vehicle to be detected based on the obtained landing points of each wheel.

[0044] Optionally, in a specific implementation, the vehicle body point cloud acquisition module includes:

[0045] The scene point cloud acquisition submodule is used to obtain the scene point cloud of the vehicle to be detected;

[0046] A depth map generation submodule is used to generate a depth map corresponding to the scene point cloud based on the ground;

[0047] A connected domain analysis submodule, configured to perform connected domain analysis on the depth map to obtain a maximum connected domain in the depth map;

[0048] The vehicle body point cloud acquisition submodule is used to acquire the point cloud located in the maximum connected domain in the scene point cloud as the vehicle body point cloud about the vehicle to be detected.

[0049] Optionally, in a specific implementation, the device further includes:

[0050] The vehicle size judgment module is used to judge whether the size of the largest connected domain meets the preset vehicle size before obtaining the point cloud located in the largest connected domain in the scene point cloud; if so, trigger the vehicle body point cloud acquisition submodule.

[0051] Optionally, in a specific implementation, the scene point cloud acquisition submodule is specifically used to:

[0052] Acquire an initial point cloud about the vehicle to be detected collected by at least one radar; and splice the acquired initial point clouds to obtain a scene point cloud about the vehicle to be detected.

[0053] Optionally, in a specific implementation, the vehicle bottom point cloud extraction module includes:

[0054] A calibration point cloud extraction submodule is used to extract point clouds located in a specified area of ​​the vehicle body point cloud as calibration point clouds;

[0055] a chassis height calculation submodule, configured to calculate the chassis height of the vehicle to be detected based on the calibration point cloud and a specified ratio matching the specified area;

[0056] The vehicle bottom point cloud sub-extraction module is used to extract the point cloud below the chassis height from the vehicle body point cloud as the vehicle bottom point cloud of the vehicle to be detected.

[0057] Optionally, in a specific implementation, the chassis height calculation submodule is specifically configured to:

[0058] Generate a height histogram of the calibration point cloud; determine, from the height histogram, each initial height whose number of point clouds satisfies a specified ratio matching the specified area; and determine the minimum height among the initial heights as the chassis height of the vehicle to be detected.

[0059] Optionally, in a specific implementation, the lower edge point cloud extraction module is specifically used to:

[0060] Based on a preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud, the vehicle bottom point cloud is clustered to obtain the position of each wheel; according to the position of each wheel, the lower edge point cloud of each wheel is extracted from the vehicle bottom point cloud.

[0061] Optionally, in a specific implementation, the vehicle posture calculation module is specifically used to:

[0062] Determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel; or,

[0063] Performing circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determining the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

[0064] Optionally, in a specific implementation, the device further includes:

[0065] The information sending module is used to send the landing point of each wheel and the target posture to a preset parking robot, so that the parking robot uses the landing point of each wheel as the optimal clamping point and clamps the vehicle to be inspected based on the target posture.

[0066] In a third aspect, an embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0067] Memory for storing computer programs;

[0068] The processor is configured to implement the steps of any vehicle posture detection method provided in the first aspect above when executing the program stored in the memory.

[0069] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the steps of any vehicle posture detection method provided in the first aspect are implemented.

[0070] In a fifth aspect, an embodiment of the present invention provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the steps of any vehicle posture detection method provided in the first aspect above.

[0071] Beneficial effects of the embodiments of the present invention:

[0072] As can be seen above, by applying the solution provided by the embodiments of the present invention, at least one radar can be installed in a parking lot, garage, or other location. When a vehicle enters the detection area of ​​each radar, each radar will collect a point cloud related to the vehicle. By stitching the point clouds collected by each radar, a scene point cloud related to the vehicle can be obtained. Furthermore, a body point cloud related to the vehicle can be extracted from this scene point cloud. This body point cloud includes at least three wheels of the vehicle.

[0073] In this way, for the vehicle to be detected, the initial point cloud obtained by collecting the vehicle to be detected by the at least one radar set up above can also be used to obtain the body point cloud of the vehicle to be detected. Then, the point cloud located between the vehicle chassis of the vehicle to be detected and the ground in the body point cloud can be extracted to obtain the bottom point cloud of the vehicle to be detected and measured. After that, the lower edge point cloud of each wheel included in the above bottom point cloud can be determined by wheel identification of the bottom point cloud. Then, the landing point of each wheel can be determined based on the bottom edge point cloud of each wheel, and the target posture of the vehicle to be detected can be calculated based on the landing points of each wheel obtained.

[0074] Based on this, by applying the solution provided by the embodiment of the present invention, the body point cloud of the vehicle to be detected can be used to accurately determine the landing points of at least three wheels of the vehicle to be detected, and then, based on the landing points of each of the above wheels, the target posture of the vehicle to be detected can be accurately calculated. In this way, when the parking robot parks the vehicle to be detected, it can use the landing point of each wheel determined above as the optimal clamping point, and clamp the vehicle to be detected based on the target posture obtained by the above calculation. In this way, after completing the clamping of the vehicle to be detected, the parking robot can further move the vehicle to be detected and complete the parking of the vehicle to be detected. In other words, by applying the solution provided by the embodiment of the present invention, the posture of the vehicle and the landing point of the wheels can be accurately obtained, the accuracy of the parking robot clamping the vehicle can be improved, and the parking robot can achieve normal parking of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present invention 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 of the present invention. For ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0076] Figure 1 A schematic diagram of a configuration of at least one radar;

[0077] Figure 2(a)-Figure 2(c) From the top view, left front view and right front view respectively Figure 1 Schematic diagram of the acquisition area of ​​the middle lidar 1-3 for observation;

[0078] Figure 3 A schematic flow chart of a vehicle posture detection method provided by an embodiment of the present invention;

[0079] Figure 4 for Figure 3 A flowchart of a specific implementation of S303;

[0080] Figure 5 A schematic flow chart of another vehicle posture detection method provided by an embodiment of the present invention;

[0081] Figure 6 A schematic diagram showing how the parking robot uses the landing point of each wheel as the optimal gripping point and grips the vehicle to be inspected based on the target pose.

[0082] Figure 7 A schematic flow chart of another vehicle posture detection method provided by an embodiment of the present invention;

[0083] Figure 8 A schematic flow chart of another vehicle posture detection method provided by an embodiment of the present invention;

[0084] Figure 9 for Figure 3 A flowchart of a specific implementation of S302;

[0085] Figure 10 for Figure 9 A flowchart of a specific implementation of S3022;

[0086] Figure 11 (a) shows a specific example. Figure 1 A schematic diagram of a scene point cloud of a vehicle to be detected obtained by stitching together the initial images of the vehicle to be detected collected by the laser radars 1-3;

[0087] FIG11( b ) is a schematic diagram of the maximum connected domain in the depth map corresponding to FIG11( a );

[0088] FIG11( c ) is a schematic diagram of the vehicle body point cloud of the point cloud to be detected in FIG11( a );

[0089] FIG11( d ) is a schematic diagram of the bottom point cloud of the vehicle to be detected in FIG11( a );

[0090] FIG11( e ) is a schematic diagram of the wheel area and landing point of each wheel in FIG11( d );

[0091] Figure 12 A schematic structural diagram of a vehicle posture detection device provided by an embodiment of the present invention;

[0092] Figure 13 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0093] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described 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 ordinary technicians in this field based on this application are within the scope of protection of the present invention.

[0094] In related art, the method for parking a vehicle by a parking robot typically involves installing a laser radar at a high point and a low point in the parking area. When a vehicle enters the parking area, the laser radar at the high point scans the vehicle to obtain its width information, while the laser radar at the low point measures the vehicle's wheel diameter and wheelbase information. The parking robot obtains the width information, wheel diameter, and wheelbase information from the two laser radars, and generates vehicle appearance information based on the width information, wheel diameter, and wheelbase information. In this way, the parking robot can clamp and carry the vehicle based on the generated vehicle appearance information to complete parking.

[0095] However, in the aforementioned related technologies, when the vehicle's direction of travel is at an angle to the laser radar's scanning area, the vehicle information obtained by the high-level laser radar and the low-level laser radar will have significant errors. This will seriously affect the parking robot's ability to accurately grasp the vehicle, resulting in the robot being unable to park the vehicle properly.

[0096] In order to solve the above technical problems, an embodiment of the present invention provides a vehicle posture detection method.

[0097] This method can be applied to any application scenario that requires the use of vehicle posture, for example, in a garage, a parking robot parks the vehicle based on the vehicle posture, or in autonomous driving, the vehicle's direction of travel is controlled based on the vehicle posture. This is all reasonable.

[0098] Furthermore, this method can be applied to various electronic devices, such as a monitoring server communicating with a radar, a radar communicating with another radar, a parking robot control server communicating with both a radar and a parking robot, and a parking robot communicating with a radar. This is all reasonable. Hereinafter, the subject executing the vehicle posture detection method provided in the embodiments of the present invention will be referred to as an electronic device.

[0099] Based on this, the embodiment of the present invention does not specifically limit the application scenario and execution subject of the method, hereinafter referred to as electronic device.

[0100] A vehicle posture detection method provided by an embodiment of the present invention may include the following steps:

[0101] Obtaining a body point cloud of the vehicle to be detected; wherein the body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar;

[0102] Extracting a bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the bottom point cloud is: a point cloud between the vehicle chassis of the vehicle to be detected and the ground;

[0103] Determining the lower edge point cloud of each wheel included in the vehicle body point cloud by wheel recognition of the vehicle bottom point cloud;

[0104] Based on the lower edge point cloud of each wheel, the landing point of each wheel is determined, and based on the obtained landing points of each wheel, the target posture of the vehicle to be detected is calculated.

[0105] As can be seen above, by applying the solution provided by the embodiments of the present invention, at least one radar can be installed in a parking lot, garage, or other location. When a vehicle enters the detection area of ​​each radar, each radar will collect a point cloud related to the vehicle. By stitching the point clouds collected by each radar, a scene point cloud related to the vehicle can be obtained. Furthermore, a body point cloud related to the vehicle can be extracted from this scene point cloud. This body point cloud includes at least three wheels of the vehicle.

[0106] In this way, for the vehicle to be detected, the initial point cloud obtained by collecting the vehicle to be detected by the at least one radar set up above can also be used to obtain the body point cloud of the vehicle to be detected. Then, the point cloud located between the vehicle chassis of the vehicle to be detected and the ground in the body point cloud can be extracted to obtain the bottom point cloud of the vehicle to be detected and measured. After that, the lower edge point cloud of each wheel included in the above bottom point cloud can be determined by wheel identification of the bottom point cloud. Then, the landing point of each wheel can be determined based on the bottom edge point cloud of each wheel, and the target posture of the vehicle to be detected can be calculated based on the landing points of each wheel obtained.

[0107] Based on this, by applying the solution provided by the embodiment of the present invention, the body point cloud of the vehicle to be detected can be used to accurately determine the landing points of at least three wheels of the vehicle to be detected, and then, based on the landing points of each of the above wheels, the target posture of the vehicle to be detected can be accurately calculated. In this way, when the parking robot parks the vehicle to be detected, it can use the landing point of each wheel determined above as the optimal clamping point, and clamp the vehicle to be detected based on the target posture obtained by the above calculation. In this way, after completing the clamping of the vehicle to be detected, the parking robot can further move the vehicle to be detected and complete the parking of the vehicle to be detected. In other words, by applying the solution provided by the embodiment of the present invention, the posture of the vehicle and the landing point of the wheels can be accurately obtained, the accuracy of the parking robot clamping the vehicle can be improved, and the parking robot can achieve normal parking of the vehicle.

[0108] Before describing in detail a vehicle posture detection method provided by an embodiment of the present invention, it is possible to first describe at least one radar used to collect an initial point cloud of a vehicle to be detected.

[0109] The at least one radar may be any type of radar capable of collecting point cloud data, for example, the radar may be a laser radar.

[0110] Depending on the specific circumstances of the actual application scenario, at least one radar can be set up in the place where the vehicle to be detected enters, for example, in a garage, parking lot, etc. Then, after the vehicle to be detected enters the place, the at least one radar can collect point clouds of the vehicle to be detected, and obtain an initial point cloud of the vehicle to be detected.

[0111] The initial point cloud collected for each vehicle to be detected includes point cloud data for at least one wheel of the vehicle to be detected. Consequently, each resulting initial point cloud can include point clouds for at least three wheels of the vehicle to be detected. These initial point clouds can then be spliced ​​together to obtain a scene point cloud for the vehicle to be detected, which can also include point clouds for at least three wheels of the vehicle to be detected.

[0112] Optionally, each radar in the at least one radar can be calibrated to the same world coordinate system, that is, the world coordinate systems of the above-mentioned each radar are the same world coordinate system, and the radar coordinate systems of the above-mentioned each radar have the same coordinate transformation relationship with the above-mentioned same world coordinate system.

[0113] In this way, for the initial point clouds of the vehicle to be detected collected by the above-mentioned radars, each point in the above-mentioned initial point clouds can be calibrated to the same world coordinate system, thereby realizing the splicing of the above-mentioned initial point clouds to obtain the scene point cloud of the vehicle to be detected.

[0114] Since each of the aforementioned initial point clouds may include point clouds of objects other than the vehicle to be detected within the acquisition range, the resulting scene point cloud related to the vehicle to be detected may also include point clouds of objects other than the vehicle to be detected. Furthermore, the point cloud related to the vehicle to be detected can be extracted from the aforementioned scene point cloud related to the vehicle to be detected, i.e., the point cloud related to the body contour of the vehicle to be detected can be extracted. The extracted point cloud is then the body point cloud related to the vehicle to be detected.

[0115] Based on this, the specific number of the at least one radar configured can optionally be the minimum number of radars required to collect an initial point cloud including the three wheels of the vehicle to be detected. This minimum number of radars can be determined based on the size of the radar site and the acquisition range of the radars used.

[0116] For example, when the size of the site where the radar is located is small and the acquisition range of the radar used is large, the specific number of the at least one radar can be 1. Thus, the initial point cloud including the three wheels of the vehicle to be detected can be collected by the one radar.

[0117] For another example, when the radar is located in a large area and the radar has a large acquisition range, the specific number of the at least one radar may be two. Thus, the two radars can be used to acquire an initial point cloud including three wheels of the vehicle to be detected. The initial point cloud acquired by one radar includes the initial point cloud for one wheel of the vehicle to be detected, while the initial point cloud acquired by the other radar includes the initial point cloud for the other two wheels of the vehicle to be detected.

[0118] For another example, when the radar site is large and the radar collection range is small, the specific number of the at least one radar may be three. Thus, the three radars can be used to collect initial point clouds including three wheels of the vehicle to be detected. The initial point cloud collected by each radar includes an initial point cloud for one wheel of the vehicle to be detected, and each initial point cloud includes a different vehicle.

[0119] It should be emphasized that the specific number of the at least one radar mentioned above, as well as the wheels and the number of wheels included in the point cloud data collected by each radar, are merely examples of setting up at least one radar in the place entered by the vehicle, and are not limiting.

[0120] For example, Figure 1 As shown, the radar is located in the garage entrance area, and the three sectors with the laser radars 1-3 as the center are the collection areas of the laser radars 1-3 respectively.

[0121] In this way, a 3D (3-Dimension) laser radar is set up in front of the storage area to scan obliquely downward. Figure 1 The laser radar 3 in the image is used to collect the two front wheels of the vehicle entering the warehouse; two 3D laser radars that scan obliquely downward are set on the right side of the warehouse area, namely Figure 1 LiDAR 1 and LiDAR 2 are used to collect data from one of the two wheels on the right side of the vehicle entering the warehouse. Therefore, the above-mentioned LiDAR lasers 1-3 collect data from the vehicle entering the warehouse respectively, thus covering all three wheels of the vehicle entering the warehouse.

[0122] That is to say, the radar lasers 1-3 respectively collect the point clouds of the vehicles entering the warehouse, and the three initial point clouds obtained include the point clouds of the three wheels of the vehicles entering the warehouse.

[0123] Among them, such as Figure 2(a)-Figure 2(c) As shown, they are respectively viewed from the top, left front and right front perspectives. Figure 1 Schematic diagram of the observation area of ​​the acquisition area of ​​the laser radar 1-3.

[0124] In this way, Figure 1 When at least one radar is set up as shown, it can stably detect information such as the vehicle's body contour, chassis height, and wheelbase without affecting the vehicle's operation, which helps to improve the intelligence of automatic parking systems including parking robots.

[0125] Furthermore, if there is no restriction on the installation height of the two 3D laser radars that scan obliquely downward on the right side of the above-mentioned storage area, and / or the collection range of the two 3D laser radars that scan obliquely downward on the right side of the above-mentioned storage area is large enough, then only one 3D laser radar can be set on the right side of the above-mentioned storage area, and the 3D laser radar can scan the two wheels on the right side of the vehicle entering the warehouse.

[0126] Below, a vehicle posture detection method provided by an embodiment of the present invention is described in detail with reference to the accompanying drawings.

[0127] Figure 3 A flow chart of a vehicle posture detection method provided by an embodiment of the present invention is shown as follows: Figure 3 As shown, the method may include the following steps S301-S304:

[0128] S301: Obtaining a body point cloud of a vehicle to be detected;

[0129] The vehicle body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing the initial point cloud of the vehicle to be detected collected by at least one radar;

[0130] When detecting the position and posture of a vehicle to be detected, a body point cloud of the vehicle to be detected may be obtained first.

[0131] As mentioned above, in actual application scenarios, at least one radar can be set up in the place where the vehicle to be detected enters, so that when the vehicle to be detected enters the collection area of ​​each radar, each radar can collect and obtain the initial point cloud about the vehicle to be detected, and each initial point cloud includes at least three wheels of the vehicle to be detected; then, by splicing each initial point cloud, a scene point cloud about the vehicle to be detected can be obtained, and the scene point cloud also includes at least three wheels of the vehicle to be detected.

[0132] In this way, a body point cloud of the vehicle to be detected can be extracted from the scene point cloud, and the body point cloud includes at least three wheels of the vehicle to be detected.

[0133] Optionally, an electronic device used to execute a vehicle posture detection method provided by an embodiment of the present invention can directly receive a body point cloud of a vehicle to be detected sent by other electronic devices.

[0134] Optionally, the electronic device may receive a scene point cloud of the vehicle to be detected sent by other electronic devices, so that the electronic device may extract a body point cloud of the vehicle to be detected from the scene point cloud.

[0135] Optionally, after collecting an initial point cloud for the vehicle to be detected, each radar may transmit the initial point cloud to the electronic device. The electronic device, upon receiving the initial point clouds, may then stitch them together to obtain a scene point cloud for the vehicle to be detected. In this way, the electronic device can extract a body point cloud for the vehicle to be detected from the scene point cloud.

[0136] Based on this, the embodiment of the present invention does not limit the specific implementation of the above step S301. For the sake of clarity, the specific implementation of the above step S301 will be described with examples later.

[0137] S302: Extracting the bottom point cloud of the vehicle to be detected from the vehicle body point cloud;

[0138] The vehicle bottom point cloud is the point cloud between the chassis of the vehicle to be detected and the ground;

[0139] After obtaining the above-mentioned body point cloud of the vehicle to be detected, the point cloud between the vehicle chassis of the vehicle to be detected and the ground can be extracted from the body point cloud, and the extracted point cloud is used as the bottom point cloud of the vehicle to be detected.

[0140] For example, the electronic device can determine the chassis height of the vehicle chassis in the body point cloud of the vehicle to be detected based on empirical values ​​of chassis heights for various types of vehicles. Furthermore, based on the determined chassis height, the corresponding position of the vehicle chassis in the body point cloud of the vehicle to be detected can be determined. In this way, the point cloud between the determined position corresponding to the vehicle chassis and the ground can be extracted to obtain a point cloud of the vehicle bottom of the vehicle to be detected.

[0141] Of course, the electronic device can also use other methods to extract the bottom point cloud of the vehicle to be detected from the vehicle body point cloud, and this embodiment of the present invention does not specifically limit this. For the sake of clarity, the specific implementation of the above step S302 will be described below with examples.

[0142] S303: determining the lower edge point cloud of each wheel included in the vehicle body point cloud by wheel recognition of the vehicle bottom point cloud;

[0143] After obtaining the bottom point cloud of the vehicle to be detected, the wheels can be identified on the bottom point cloud to determine the partial edge point clouds of each wheel included in the vehicle body point cloud.

[0144] Taking into account that part of the vehicle's wheels are located above the vehicle's chassis and part are located below the vehicle's chassis, and the bottom point cloud of the vehicle to be detected is: the point cloud between the vehicle chassis of the vehicle to be detected and the ground in the body point cloud of the vehicle to be detected, therefore, the above-mentioned wheel recognition is performed on the bottom point cloud, and the partial edge point cloud of each wheel included in the above-mentioned body point cloud is: the lower edge point cloud of each wheel included in the above-mentioned body point cloud.

[0145] Optionally, in a specific implementation, such as Figure 4 As shown, the above step S303 may include the following steps S3031-S3032:

[0146] S3031: Based on a preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud, cluster the vehicle bottom point cloud to obtain the position of each wheel;

[0147] S3032: Extracting the lower edge point cloud of each wheel from the vehicle bottom point cloud according to the position of each wheel.

[0148] In this specific implementation, in actual application scenarios, the types of vehicles entering parking lots, garages, and other places are generally well-defined, and thus the wheel sizes of vehicles entering these places fall within a relatively well-defined wheel size range. For example, a garage in a residential area typically receives family cars; another example is a shopping mall parking lot, where vehicles typically enter include family cars and smaller trucks used by the mall.

[0149] In addition, based on information related to the vehicle structure, the wheels of the vehicle entering the above-mentioned place have a certain positional relationship, and the positional relationship includes: the line connecting the two front wheels of the vehicle is parallel to the line connecting the two rear wheels of the vehicle, the line connecting the two wheels on the right side of the vehicle is parallel to the center of the two wheels on the left side of the vehicle, the line connecting the two wheels on the right side of the vehicle is perpendicular to the line connecting the two front wheels of the vehicle and the line connecting the two rear wheels of the vehicle, and the line connecting the two wheels on the left side of the vehicle is perpendicular to the line connecting the two front wheels of the vehicle and the line connecting the two rear wheels of the vehicle.

[0150] Once the underbody point cloud of the vehicle to be inspected is obtained, it can be clustered based on the preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud to determine the position of each wheel. Furthermore, based on the determined wheel positions, the lower edge point cloud of each wheel can be extracted from the underbody point cloud of the vehicle to be inspected.

[0151] S304: Determine the landing point of each wheel based on the lower edge point cloud of each wheel, and calculate the target pose of the vehicle to be detected based on the obtained landing points of each wheel.

[0152] After determining the lower edge point cloud of each wheel included in the above-mentioned body point cloud of the vehicle to be detected, the landing point of each wheel can be determined based on the lower edge point cloud of each wheel. Then, according to the positional relationship between each wheel, the target posture of the vehicle to be detected can be calculated based on the obtained landing points of each wheel.

[0153] Optionally, in a specific implementation, in the above step S304, determining the landing point of each wheel based on the lower edge point cloud of each wheel may include the following step 11:

[0154] Step 11: Determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel.

[0155] In this specific implementation, after obtaining the lower edge point cloud of each wheel included in the body point cloud of the vehicle to be detected, the intersection of the lower edge point cloud of each wheel and the ground can be directly determined, and the determined intersection points are the landing points of each wheel.

[0156] Optionally, in a specific implementation, in the above step S304, determining the landing point of each wheel based on the lower edge point cloud of each wheel may include the following step 21:

[0157] Step 21: Perform circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determine the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

[0158] In this specific implementation, considering that the wheels of the vehicle to be inspected may not be aligned, in order to improve the accuracy of the determined touchdown point of each wheel and, consequently, the accuracy of the ultimately determined target pose of the vehicle to be inspected, after obtaining the lower edge point cloud of each wheel included in the body point cloud of the vehicle to be inspected, a circular fit is first performed on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel. This allows the intersection of each wheel area with the ground to be determined, and the center point of each determined intersection is then determined as the touchdown point of each wheel.

[0159] In this way, the problem of low accuracy in target posture calculation caused by the wheels of the vehicle to be detected not being straightened can be solved, thereby improving the accuracy of the landing point of each wheel determined, and further improving the accuracy of the target posture of the vehicle to be detected finally determined.

[0160] Optionally, in a specific implementation, such as Figure 5 As shown, a vehicle posture detection method provided by an embodiment of the present invention may further include the following step S305:

[0161] S305: Sending the landing point of each wheel and the target posture to a preset parking robot, so that the parking robot uses the landing point of each wheel as the optimal clamping point and parks the vehicle to be inspected based on the target posture.

[0162] In this specific implementation, a parking robot can be used to park the vehicle to be inspected. In order to improve the accuracy of the parking robot in clamping the vehicle and enable the parking robot to park the vehicle normally, a vehicle posture detection method provided in an embodiment of the present invention can be used to obtain the landing point and target posture of each wheel of the vehicle to be inspected.

[0163] The above-determined contact points of each wheel are: the contact points of each wheel included in the acquired body point cloud of the vehicle to be detected.

[0164] In this way, after obtaining the landing point and target posture of each wheel of the vehicle to be detected, the landing point of each wheel and the target posture of the vehicle to be detected can be sent to the preset parking robot. Figure 6 As shown, after receiving the landing point and target position of each wheel, the parking robot uses the landing point as the optimal gripping point and, based on the target position of the vehicle to be inspected, uses its gripping arms to grip each wheel. The parking robot then maneuvers the vehicle to a designated vacant parking space, completing the parking process.

[0165] Optionally, when the executor of a vehicle posture detection method provided by an embodiment of the present invention is the above-mentioned parking robot, the parking robot can directly use the landing point of each wheel as the optimal clamping point after obtaining the landing point of each wheel and the target posture of the vehicle to be detected, and based on the target posture of the vehicle to be detected, use the clamping arm to clamp each wheel to complete the parking of the vehicle to be detected.

[0166] As can be seen from the above, by applying the solution provided by the embodiment of the present invention, the body point cloud of the vehicle to be detected can be used to accurately determine the landing points of at least three wheels of the vehicle to be detected, and then, based on the landing points of each wheel, the target posture of the vehicle to be detected can be accurately calculated. In this way, when parking the vehicle to be detected, the parking robot can use the landing point of each wheel determined above as the optimal clamping point, and clamp the vehicle to be detected based on the target posture obtained by the above calculation. In this way, after completing the clamping of the vehicle to be detected, the parking robot can further move the vehicle to be detected and complete the parking of the vehicle to be detected. In other words, by applying the solution provided by the embodiment of the present invention, the posture of the vehicle and the landing point of the wheel can be accurately obtained, the accuracy of the parking robot clamping the vehicle can be improved, and the parking robot can achieve normal parking of the vehicle.

[0167] Optionally, in a specific implementation, such as Figure 7 As shown, the above step S301, obtaining the body point cloud of the vehicle to be detected, may include the following steps S3011-S3014:

[0168] S3011: Obtaining a scene point cloud of the vehicle to be detected;

[0169] S3012: Generate a depth map corresponding to the scene point cloud using the ground as a reference plane;

[0170] S3013: Performing connected domain analysis on the depth map to obtain the maximum connected domain in the depth map;

[0171] S3014: Extracting the point cloud in the area corresponding to the largest connected domain from the scene point cloud as the body point cloud of the vehicle to be detected.

[0172] In this specific implementation, the electronic device may first obtain a scene point cloud related to the vehicle to be detected. This scene point cloud is obtained by stitching together initial point clouds related to the vehicle to be detected collected by at least one radar. Since each initial point cloud includes point clouds of at least three wheels of the vehicle to be detected, the obtained scene point cloud related to the vehicle to be detected also includes point clouds of at least three wheels of the vehicle to be detected.

[0173] Optionally, the above step S3011 may include the following steps 31-32:

[0174] Step 31: Obtaining an initial point cloud of the vehicle to be detected collected by at least one radar;

[0175] Step 32: stitching the acquired initial point clouds to obtain a scene point cloud of the vehicle to be detected.

[0176] In the embodiment shown in steps 31-32 above, after each of the at least one radar collects an initial point cloud for the vehicle to be detected, the collected initial point cloud can be sent to the electronic device. The electronic device can then obtain each initial point cloud for the vehicle to be detected. Furthermore, the initial point clouds can be spliced ​​together to obtain a scene point cloud for the vehicle to be detected. Furthermore, since each initial point cloud includes point clouds for at least three wheels of the vehicle to be detected, the scene point cloud can also include point clouds for at least three wheels of the vehicle to be detected.

[0177] Optionally, each radar in the at least one radar can be calibrated to the same world coordinate system, that is, the world coordinate systems of the above-mentioned each radar are the same world coordinate system, and the radar coordinate systems of the above-mentioned each radar have the same coordinate transformation relationship with the above-mentioned same world coordinate system.

[0178] In this way, for the initial point clouds of the vehicle to be detected collected by the above-mentioned radars, the electronic device can calibrate each point in the above-mentioned initial point clouds to the same world coordinate system, thereby realizing the splicing of the above-mentioned initial point clouds to obtain the scene point cloud of the vehicle to be detected.

[0179] After obtaining the scene point cloud of the vehicle to be detected, a depth map corresponding to the scene point cloud can be generated using the ground as the reference plane. This allows a connected domain analysis to be performed on the depth map to determine each connected domain in the depth map and the largest connected domain among the connected domains. In other words, the largest connected domain in the depth map can be obtained by performing a connected domain analysis on the depth map.

[0180] Among them, since usually, in actual application scenarios, the connected domain of the vehicle to be detected is the largest connected domain among the above-mentioned connected domains within the scope of the place entered by the vehicle to be detected, the edge contour of the largest connected domain in the above-mentioned depth map can be used as the body contour of the vehicle to be detected.

[0181] Based on this, the maximum connected domain in the depth map can be remapped back to the scene point cloud of the vehicle to be detected to obtain the area in the scene point cloud corresponding to the maximum connected domain in the depth map. Furthermore, the point cloud in the scene point cloud corresponding to the maximum connected domain in the depth map can be extracted and used as the body point cloud of the vehicle to be detected.

[0182] Typically, in practical applications, the types of vehicles entering parking lots, garages, and other locations are relatively well-defined, and therefore, the sizes of vehicles entering these locations fall within a relatively well-defined size range. For example, a garage in a residential complex typically hosts family cars; another example is a shopping mall parking lot, where vehicles typically enter include family cars and smaller trucks used by the mall.

[0183] That is to say, since the edge contour of the largest connected domain in the depth map can be used as the body contour of the vehicle to be detected, the size of the largest connected domain in the depth map can be within the vehicle size range that matches the actual application scenario.

[0184] Based on this, optionally, in a specific implementation, such as Figure 8 As shown, the vehicle posture detection method provided by the above embodiment of the present invention may further include the following step S3015:

[0185] S3015: Determine whether the size of the largest connected domain falls within a preset vehicle size range; if so, execute the above step S3014.

[0186] In this specific implementation, after obtaining the maximum connected domain in the depth map, it can be first determined whether the size of the maximum connected domain meets a preset vehicle size range.

[0187] The above-mentioned preset vehicle size range may be determined based on the types of vehicles entering parking lots, garages, and other places in actual application scenarios.

[0188] Alternatively, in actual application scenarios, the types of vehicles entering parking lots, garages, and other places can be counted, and then the sizes of each type of vehicle can be found. In this way, the minimum size among the found sizes can be used as the lower limit of the preset vehicle size range, and the maximum size among the found sizes can be used as the upper limit of the preset vehicle size range, thereby obtaining the preset vehicle size range.

[0189] In this way, when the judgment result is yes, that is, when the size of the largest connected domain in the above depth map belongs to the preset vehicle size range, it can be explained that the edge contour of the largest connected domain can be used as the body contour of the vehicle to be detected. Thus, the point cloud in the area corresponding to the largest connected domain in the scene point cloud can be extracted as the body point cloud of the vehicle to be detected.

[0190] Optionally, when the above judgment result is no, that is, the size of the largest connected domain in the above depth map does not belong to the preset vehicle size range, it can be explained that the edge contour of the largest connected domain cannot be used as the body contour of the vehicle to be detected, and then, it can be determined that the object entering the above place is not a vehicle, and thus, the execution of the above vehicle posture detection method can be stopped.

[0191] Optionally, when the above judgment result is no, that is, the size of the maximum connected domain in the above depth map does not belong to the preset vehicle size range, then considering that there may be a deviation in the maximum connected domain size due to a small vehicle entry distance, it can be explained that the edge contour of the maximum connected domain may or may not be used as the body contour of the vehicle to be detected.

[0192] In this way, data of the maximum connected domain whose size exceeds the preset vehicle size range can be calculated first.

[0193] Among them, when the data is less than the preset threshold, it can be considered that the edge contour of the maximum connected domain can be used as the body contour of the vehicle to be detected, thereby extracting the point cloud located in the area corresponding to the maximum connected domain in the scene point cloud as the body point cloud of the vehicle to be detected.

[0194] When the data is not less than a preset threshold, it can be considered that the edge contour of the maximum connected domain cannot be used as the body contour of the vehicle to be detected, and thus the above-mentioned vehicle posture detection method is stopped.

[0195] Optionally, in a specific implementation, such as Figure 9As shown, the above step S302, extracting the bottom point cloud of the vehicle to be detected from the vehicle body point cloud, may include the following steps S3021-S3023:

[0196] S3021: extracting the point cloud located in the specified area of ​​the vehicle body point cloud as the calibration point cloud;

[0197] S3022: Determine the chassis height of the vehicle to be detected based on the calibration point cloud and the specified ratio matching the specified area;

[0198] S3023: Extract the point cloud below the chassis height from the vehicle body point cloud as the bottom point cloud of the vehicle to be detected.

[0199] In this specific implementation, for each position of the vehicle to be detected, the ratio of the vehicle's chassis height to the height of the position is fixed. However, for different positions of the vehicle to be detected, the ratio of the vehicle's chassis height to the height of the position can be the same or different.

[0200] In this way, for the vehicle to be detected, the ratio of the chassis height of the vehicle to be detected to the height of the position at each position of the vehicle to be detected can be predetermined.

[0201] Based on this, for each point cloud region of the vehicle body, the ratio of the number of points located between the vehicle chassis and the ground to the number of points in that point cloud region is also fixed. For different point cloud regions in the vehicle body point cloud, the ratio of the number of points located between the vehicle chassis and the ground to the number of points in that point cloud region can be the same or different.

[0202] For the sake of convenience, the ratio of the chassis height of the vehicle to be detected to the height of each position of the vehicle to be detected can be referred to as the first ratio; and the ratio of the number of points between the chassis of the vehicle to be detected and the ground in the point cloud area to the number of points in the point cloud area for each point cloud area in the body point cloud of the vehicle to be detected can be referred to as the second ratio.

[0203] In this way, considering that the body point cloud of the vehicle to be detected can reflect the appearance of the vehicle to be detected, for each position of the vehicle to be detected, the above-mentioned first ratio is correlated with the second ratio of the point cloud area corresponding to the position in the body point cloud of the vehicle to be detected.

[0204] Based on this, when performing posture detection on a vehicle to be detected, a designated area in the body point cloud of the vehicle to be detected and a designated ratio matching the designated area can be determined first.

[0205] For example, the designated area is an area located in the middle of the vehicle body point cloud and has a width of 20 centimeters. The designated ratio of the designated area is 5%.

[0206] In this way, after obtaining the body point cloud of the vehicle to be detected, the point cloud located in the specified area of ​​the body point cloud can be extracted as the calibration point cloud. Then, based on the specified ratio between the extracted calibration point cloud and the specified area, the chassis height of the vehicle to be detected can be determined.

[0207] Optional, such as Figure 10 As shown, the above step S3022 may include the following steps S3022A-S3022C:

[0208] S3022A: Generate height histogram of calibration point cloud;

[0209] S3022B: Based on the height histogram, determine each height interval where the calibration ratio is equal to the specified ratio matching the specified area;

[0210] The calibration ratio of each height interval is: the ratio of the number of points included in the height interval to the number of points included in the designated area;

[0211] S3022C: Determine the upper limit value of the minimum height interval in each height interval as the chassis height of the vehicle to be detected.

[0212] In the above Figure 7 In the embodiment shown, considering that there are noise points and miscellaneous points in the point cloud extracted between the vehicle chassis of the vehicle to be detected and the ground, in order to remove as many noise points and miscellaneous points as possible in the point cloud extracted from the bottom of the vehicle to be detected and improve the accuracy of the determined chassis height of the vehicle to be detected, the chassis height of the vehicle to be detected can be determined by means of a height histogram.

[0213] Based on this, after extracting the calibration point cloud, a height histogram of the calibration point cloud can be generated. Furthermore, the ratio of the number of points included in each height interval in the height histogram to the number of points included in the designated area can be calculated as the calibration ratio for that height interval. In this way, based on the height histogram, each height interval whose calibration ratio equals the designated ratio matching the designated area can be determined. Subsequently, the minimum height interval among the initially determined height intervals can be selected, and the upper limit of this minimum height interval can be determined as the chassis height of the vehicle to be detected.

[0214] For example, the designated area is located in the middle of the vehicle point cloud and has a width of 20 cm. The designated ratio for matching the designated area is 5%. Consequently, a height histogram with a height range of 0-3 meters and a 5 mm interval can be generated. From this height histogram, the height intervals with a 5% calibration ratio can be determined. The upper limit of the minimum height interval among these determined height intervals is then determined as the chassis height of the vehicle to be detected.

[0215] In this way, after obtaining the chassis height of the above-mentioned vehicle to be detected, the point cloud below the calculated chassis height can be extracted from the above-mentioned body point cloud of the above-mentioned vehicle to be detected, and the extracted point cloud can be used as the bottom point cloud of the above-mentioned vehicle to be detected.

[0216] In order to facilitate understanding of the vehicle posture detection method provided by the above embodiment of the present invention, the following Figure 1 Based on the setting method of at least one radar shown, a vehicle posture detection method provided by the above embodiment of the present invention is specifically described through a specific example.

[0217] As shown in Figure 11(a), Figure 1 The scene point cloud of the vehicle to be detected is obtained by stitching together the initial images of the vehicle to be detected collected by the laser radars 1-3.

[0218] Then, the depth map corresponding to FIG11(a) can be generated with the ground as the reference screen, and the connected domain analysis can be performed on the depth map to obtain the maximum connected domain in the depth map, as shown in the gray area in FIG11(b).

[0219] When the maximum connected domain shown in Figure 11(b) belongs to the preset vehicle size range, the point cloud in the area corresponding to the maximum connected domain shown in Figure 11(b) in the scene point cloud shown in Figure 11(a) can be extracted to obtain the vehicle body point cloud regarding the point cloud to be detected, as shown inside the stereoscopic frame in Figure 11(c).

[0220] From the vehicle body point cloud for the point cloud to be detected shown in Figure 11(c), a 20 cm wide area located in the center of the vehicle body point cloud is extracted as the designated area, and a designated ratio of 5% is determined for the designated area. Furthermore, a height histogram with a height range of 0-3 meters and a 5 mm interval can be generated. Thus, the height intervals with a calibrated ratio equal to 5% can be determined from the height histogram. The upper limit of the minimum height interval among the determined height intervals is then determined as the chassis height of the vehicle to be detected.

[0221] In this way, the point cloud below the chassis height can be extracted from the body point cloud of the vehicle to be detected as the bottom point cloud of the vehicle to be detected, as shown in the area formed by the white lines and white squares in Figure 11(d).

[0222] Furthermore, based on the preset wheel size range and the positional relationship of each wheel included in the body point cloud of the vehicle to be detected, the bottom point cloud shown in FIG11( d ) is clustered to obtain the positions of each of the above wheels; and according to the positions of each of the above wheels, the lower edge point clouds of each of the above wheels are extracted from the bottom point cloud shown in FIG11( d ), as shown in the white boxes in FIG11( d ).

[0223] After that, a circular fitting can be performed on the lower edge point cloud of each wheel as shown in FIG11(e) to obtain the wheel area of ​​each wheel, as shown in the circles in FIG11(e); and the center point of the intersection of the wheel area of ​​each wheel and the ground is determined as the landing point of each wheel, as shown in the point where the bottom of each circle in FIG11(e) is connected to the white straight line.

[0224] Corresponding to the vehicle posture detection method provided by the above-mentioned embodiment of the present invention, the embodiment of the present invention also provides a vehicle posture detection device.

[0225] Figure 12 A schematic diagram of the structure of a vehicle posture detection device provided by an embodiment of the present invention is shown in FIG. Figure 12 As shown, the device may include the following modules:

[0226] The vehicle body point cloud acquisition module 1201 is configured to acquire a vehicle body point cloud of the vehicle to be detected. The vehicle body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected. The scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar.

[0227] The vehicle bottom point cloud extraction module 1202 is used to extract the vehicle bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the vehicle bottom point cloud is: the point cloud between the vehicle chassis of the vehicle to be detected and the ground;

[0228] A lower edge point cloud extraction module 1203 is configured to determine the lower edge point cloud of each wheel included in the vehicle body point cloud by identifying the wheels in the vehicle bottom point cloud;

[0229] The vehicle posture calculation module 1204 is used to determine the landing point of each wheel based on the lower edge point cloud of each wheel, and calculate the target posture of the vehicle to be detected based on the obtained landing points of each wheel.

[0230] As can be seen above, by applying the solution provided by the embodiments of the present invention, at least one radar can be installed in a parking lot, garage, or other location. When a vehicle enters the detection area of ​​each radar, each radar will collect a point cloud related to the vehicle. By stitching the point clouds collected by each radar, a scene point cloud related to the vehicle can be obtained. Furthermore, a body point cloud related to the vehicle can be extracted from this scene point cloud. This body point cloud includes at least three wheels of the vehicle.

[0231] In this way, for the vehicle to be detected, the initial point cloud obtained by collecting the vehicle to be detected by the at least one radar set up above can also be used to obtain the body point cloud of the vehicle to be detected. Then, the point cloud located between the vehicle chassis of the vehicle to be detected and the ground in the body point cloud can be extracted to obtain the bottom point cloud of the vehicle to be detected and measured. After that, the lower edge point cloud of each wheel included in the above bottom point cloud can be determined by wheel identification of the bottom point cloud. Then, the landing point of each wheel can be determined based on the bottom edge point cloud of each wheel, and the target posture of the vehicle to be detected can be calculated based on the landing points of each wheel obtained.

[0232] Based on this, by applying the solution provided by the embodiment of the present invention, the body point cloud of the vehicle to be detected can be used to accurately determine the landing points of at least three wheels of the vehicle to be detected, and then, based on the landing points of each of the above wheels, the target posture of the vehicle to be detected can be accurately calculated. In this way, when the parking robot parks the vehicle to be detected, it can use the landing point of each wheel determined above as the optimal clamping point, and clamp the vehicle to be detected based on the target posture obtained by the above calculation. In this way, after completing the clamping of the vehicle to be detected, the parking robot can further move the vehicle to be detected and complete the parking of the vehicle to be detected. In other words, by applying the solution provided by the embodiment of the present invention, the posture of the vehicle and the landing point of the wheels can be accurately obtained, the accuracy of the parking robot clamping the vehicle can be improved, and the parking robot can achieve normal parking of the vehicle.

[0233] Optionally, in a specific implementation, the vehicle body point cloud acquisition module 1201 includes:

[0234] The scene point cloud acquisition submodule is used to obtain the scene point cloud of the vehicle to be detected;

[0235] A depth map generation submodule is used to generate a depth map corresponding to the scene point cloud based on the ground;

[0236] A connected domain analysis submodule, configured to perform connected domain analysis on the depth map to obtain a maximum connected domain in the depth map;

[0237] The vehicle body point cloud acquisition submodule is used to acquire the point cloud located in the maximum connected domain in the scene point cloud as the vehicle body point cloud about the vehicle to be detected.

[0238] Optionally, in a specific implementation, the device further includes:

[0239] The vehicle size judgment module is used to judge whether the size of the largest connected domain meets the preset vehicle size before obtaining the point cloud located in the largest connected domain in the scene point cloud; if so, trigger the vehicle body point cloud acquisition submodule.

[0240] Optionally, in a specific implementation, the scene point cloud acquisition submodule is specifically used to:

[0241] Acquire an initial point cloud about the vehicle to be detected collected by at least one radar; and splice the acquired initial point clouds to obtain a scene point cloud about the vehicle to be detected.

[0242] Optionally, in a specific implementation, the vehicle bottom point cloud extraction module 1202 includes:

[0243] A calibration point cloud extraction submodule is used to extract point clouds located in a specified area of ​​the vehicle body point cloud as calibration point clouds;

[0244] a chassis height calculation submodule, configured to calculate the chassis height of the vehicle to be detected based on the calibration point cloud and a specified ratio matching the specified area;

[0245] The vehicle bottom point cloud sub-extraction module is used to extract the point cloud below the chassis height from the vehicle body point cloud as the vehicle bottom point cloud of the vehicle to be detected.

[0246] Optionally, in a specific implementation, the chassis height calculation submodule is specifically configured to:

[0247] Generate a height histogram of the calibration point cloud; determine, from the height histogram, each initial height whose number of point clouds satisfies a specified ratio matching the specified area; and determine the minimum height among the initial heights as the chassis height of the vehicle to be detected.

[0248] Optionally, in a specific implementation, the lower edge point cloud extraction module 1203 is specifically configured to:

[0249] Based on a preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud, the vehicle bottom point cloud is clustered to obtain the position of each wheel; according to the position of each wheel, the lower edge point cloud of each wheel is extracted from the vehicle bottom point cloud.

[0250] Optionally, in a specific implementation, the vehicle posture calculation module 1204 is specifically configured to:

[0251] Determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel; or,

[0252] Performing circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determining the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

[0253] Optionally, in a specific implementation, the device further includes:

[0254] The information sending module is used to send the landing point of each wheel and the target posture to a preset parking robot, so that the parking robot uses the landing point of each wheel as the optimal clamping point and clamps the vehicle to be inspected based on the target posture.

[0255] Corresponding to the vehicle posture detection method provided by the above embodiment of the present invention, the embodiment of the present invention also provides an electronic device, such as Figure 13 As shown, it includes a processor 1301, a communication interface 1302, a memory 1303 and a communication bus 1304, wherein the processor 1301, the communication interface 1302, and the memory 1303 communicate with each other through the communication bus 1304.

[0256] Memory 1303, used for storing computer programs;

[0257] The processor 1301 is configured to implement the steps of any vehicle posture detection method provided by the above-mentioned embodiments of the present invention when executing the program stored in the memory 1303.

[0258] The communication bus mentioned in the electronic device mentioned above may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is used in the figure, but this does not mean that there is only one bus or only one type of bus.

[0259] The communication interface is used for communication between the above electronic device and other devices.

[0260] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0261] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0262] In another embodiment provided by the present invention, a computer-readable storage medium is also provided, which stores a computer program. When the computer program is executed by a processor, the steps of any vehicle posture detection method provided by the above-mentioned embodiment of the present invention are implemented.

[0263] In another embodiment provided by the present invention, a computer program product containing instructions is also provided, which, when executed on a computer, enables the computer to execute the steps of any vehicle posture detection method provided by the above-mentioned embodiments of the present invention.

[0264] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0265] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0266] Each embodiment in this specification is described in a related manner. Similar portions between the various embodiments can be referenced to each other. Each embodiment focuses on the differences between the other embodiments. In particular, since the apparatus embodiments, electronic device embodiments, computer-readable storage medium embodiments, and computer program product embodiments are generally similar to the method embodiments, their descriptions are relatively simple. For related portions, reference can be made to the descriptions of the method embodiments.

[0267] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A vehicle posture detection method, characterized in that: The method comprises: Obtaining a body point cloud of the vehicle to be detected; wherein the body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar; the at least one radar is installed in a location entered by the vehicle to be detected; Extracting a bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the bottom point cloud is: a point cloud between the vehicle chassis of the vehicle to be detected and the ground; Determining the lower edge point cloud of each wheel included in the vehicle body point cloud by wheel recognition of the vehicle bottom point cloud; Determining the landing point of each wheel based on the lower edge point cloud of each wheel, and calculating the target pose of the vehicle to be detected based on the obtained landing points of each wheel; The step of determining the landing point of each wheel based on the lower edge point cloud of each wheel includes: Determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel; or, Performing circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determining the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

2. The method according to claim 1, characterized in that The step of obtaining a body point cloud of the vehicle to be detected includes: Obtain scene point cloud of the vehicle to be detected; Taking the ground as a reference plane, generating a depth map corresponding to the scene point cloud; Performing a connected domain analysis on the depth map to obtain a maximum connected domain in the depth map; A point cloud located in an area corresponding to the largest connected domain in the scene point cloud is extracted as a body point cloud of the vehicle to be detected.

3. The method according to claim 2, characterized in that Before extracting the point cloud located in the area corresponding to the maximum connected component from the scene point cloud, the method further includes: Determining whether the size of the maximum connected domain falls within a preset vehicle size range; If so, extract the point cloud in the scene point cloud that is located in the area corresponding to the largest connected domain.

4. The method according to claim 2, characterized in that The step of obtaining a scene point cloud of a vehicle to be detected includes: Obtaining an initial point cloud of a vehicle to be detected collected by at least one radar; The obtained initial point clouds are spliced ​​to obtain the scene point cloud of the vehicle to be detected.

5. The method according to claim 1, wherein Extracting the bottom point cloud of the vehicle to be detected from the vehicle body point cloud includes: Extracting a point cloud located in a specified area from the vehicle body point cloud as a calibration point cloud; Determining a chassis height of the vehicle to be detected based on a specified ratio of the calibration point cloud and the specified area; From the vehicle body point cloud, the point cloud below the chassis height is extracted as the bottom point cloud of the vehicle to be detected.

6. The method according to claim 5, characterized in that The calculating the chassis height of the vehicle to be detected based on the calibration point cloud and the specified ratio matching the specified area includes: Generating a height histogram of the calibration point cloud; Based on the height histogram, determining each height interval having a calibration ratio equal to a specified ratio matching the specified area; wherein the calibration ratio of each height interval is: a ratio of the number of points included in the height interval to the number of points included in the specified area; The upper limit value of the minimum height interval in each height interval is determined as the chassis height of the vehicle to be detected.

7. The method according to claim 1, characterized in that The step of determining the lower edge point cloud of each wheel included in the vehicle body point cloud by wheel recognition of the vehicle bottom point cloud comprises: Based on a preset wheel size range and the positional relationship of each wheel included in the vehicle body point cloud, clustering the vehicle bottom point cloud to obtain the position of each wheel; According to the position of each wheel, the lower edge point cloud of each wheel is extracted from the vehicle bottom point cloud.

8. The method according to any one of claims 1 to 7, characterized in that The method further comprises: The landing point of each wheel and the target posture are sent to a preset parking robot, so that the parking robot uses the landing point of each wheel as the optimal clamping point and parks the vehicle to be detected based on the target posture.

9. A vehicle posture detection device, characterized in that: The device comprises: A vehicle body point cloud acquisition module, configured to acquire a vehicle body point cloud of the vehicle to be detected; wherein the vehicle body point cloud is extracted from a scene point cloud of the vehicle to be detected and includes at least three wheels of the vehicle to be detected; the scene point cloud is obtained by splicing initial point clouds of the vehicle to be detected collected by at least one radar; the at least one radar is installed in the location entered by the vehicle to be detected; A vehicle bottom point cloud extraction module is used to extract the vehicle bottom point cloud of the vehicle to be detected from the vehicle body point cloud; wherein the vehicle bottom point cloud is: the point cloud between the vehicle chassis of the vehicle to be detected and the ground; a lower edge point cloud extraction module, configured to determine the lower edge point cloud of each wheel included in the vehicle body point cloud by identifying the wheels of the vehicle bottom point cloud; A vehicle posture calculation module is used to determine the landing point of each wheel based on the lower edge point cloud of each wheel, and calculate the target posture of the vehicle to be detected based on the obtained landing points of each wheel; The vehicle posture calculation module is specifically used to determine the intersection of the lower edge point cloud of each wheel and the ground as the landing point of each wheel; or, to perform circular fitting on the lower edge point cloud of each wheel to obtain the wheel area of ​​each wheel, and determine the center point of the intersection of the wheel area of ​​each wheel and the ground as the landing point of each wheel.

10. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory for storing computer programs; A processor, configured to implement the method steps described in any one of claims 1 to 8 when executing a program stored in a memory.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps of any one of claims 1 to 8 are implemented.

Images