A space parking space recognition method, system and automobile

Abstract

The application provides a kind of space parking space identification method, system and car, comprising, by vehicle-mounted radar, the point cloud data of the environment around the car is obtained, and classification is carried out, and ground point cloud data and ground above point cloud data are obtained;The ground point cloud data is clustered and processed to identify ground obstacles, and determine the available parking space according to the identified ground obstacles;According to the available parking space, the wall surface point cloud data in the ground above point cloud data is eliminated, and the elimination result is clustered and processed to identify the suspended obstacle in the available parking space;According to the suspended obstacle in the available parking space, the lowest height value of the suspended obstacle is determined, and is combined with the available parking space, and the final space parking space is output.The application realizes accurate judgment for suspended obstacle in the case of parking space shortage scene in underground parking lot, and further achieves wide range retrieval and output for three-dimensional available space parking space.

Description

Technical Field

[0001] This invention relates to the field of spatial parking space recognition technology, and in particular to a spatial parking space recognition method, system, and vehicle. Background Technology

[0002] With the advancement of autonomous driving technology and the accelerating pace of global urbanization in recent years, vehicle intelligent systems have become crucial software infrastructure for improving the convenience and efficiency of daily life. Among these, automatic parking assistance systems are a significant application of active collision avoidance in low-speed, complex urban environments, and have become a research hotspot in recent years. Parking space detection is a critical component of automatic parking assistance systems. By analyzing and extrapolating parking space information measured by sensors (such as cameras, ultrasonic radar, and lidar), real-time detection of available parking spaces can be achieved. In contrast, the construction of parking lots has been relatively slow, and the problem of parking difficulties is becoming increasingly prominent. Researching methods for parking space detection can not only effectively improve parking space utilization and alleviate the problem of limited parking resources, but also effectively meet the requirements of parking lots in terms of efficiency, safety, and management.

[0003] Currently, there are many methods for parking space detection, which can be broadly categorized into multi-sensor collaboration between the site and vehicles, and single-vehicle sensor methods, based on their working principles. However, for multi-sensor collaboration, the high cost of building environmental perception infrastructure and the need for supporting complete fusion algorithms significantly limit the application scenarios of this method. Furthermore, the tight coupling and fusion of multiple sensors remains a challenging problem that has not yet been well resolved. For single-sensor methods, ultrasonic and lidar are commonly used, unaffected by close-range lighting conditions. Data processing is direct and therefore fast, typically real-time. Thus, in most scenarios, it meets the requirements for parking measurement in terms of measurement distance and accuracy. However, it also has some drawbacks. For example, ultrasonic radar suffers from large beam angles, poor directionality, low resolution, and short working distances; while lidar operation is greatly affected by weather and atmospheric conditions, with attenuation increasing sharply in rainy and foggy weather, significantly impacting its propagation distance. Summary of the Invention

[0004] The purpose of this invention is to propose a spatial parking space recognition method, system, and vehicle, which enables accurate identification of suspended obstacles in scenarios where parking spaces are scarce in underground parking lots, thereby achieving a wide-range retrieval of available three-dimensional parking spaces and outputting high-precision information on available parking spaces.

[0005] On the one hand, a spatial parking space recognition method is provided, including:

[0006] The vehicle acquires point cloud data of the surrounding environment through vehicle-mounted radar, and classifies the point cloud data of the surrounding environment to obtain ground point cloud data and above-ground point cloud data.

[0007] Clustering is performed on the ground point cloud data to identify ground obstacles, and available parking spaces are determined based on the identified ground obstacles.

[0008] Based on the available parking space, the wall point cloud data in the above-ground point cloud data is removed, and the removal results are clustered to identify suspended obstacles in the available parking space.

[0009] The minimum height of the suspended obstacle is determined based on the available parking space, and the minimum height of the suspended obstacle is used as the height of the available parking space and combined with the available parking space to output the final parking space.

[0010] Preferably, the classification of point cloud data specifically includes:

[0011] Ground points in the point cloud data are identified by a preset point recognition rule, and the plane where the ground points are located is taken as the ground plane;

[0012] The ground plane is calibrated using a preset transformation matrix to obtain ground point cloud data;

[0013] The point cloud data excluding the ground point cloud data is output as above-ground point cloud data.

[0014] Preferably, the step of clustering the ground point cloud data to identify ground obstacles specifically includes:

[0015] Identify the point cloud information of the target obstacle in the ground point cloud data, and segment the point cloud information of the target obstacle according to a preset standard to obtain the segmentation result;

[0016] Cluster each segment in the segmentation result according to the preset minimum and maximum number of cluster points, and output the corresponding clustering results as ground obstacles.

[0017] Preferably, determining the available parking space based on identified ground obstacles specifically includes:

[0018] Determine whether the parking space of the target parking space meets the preset parking standards based on the extent of ground obstacles;

[0019] If the parking space of the target parking space meets the preset parking standards, then the parking space is determined to be a usable parking space.

[0020] If the parking space of the target parking space does not meet the preset parking standards, the parking space is determined to be an unusable parking space.

[0021] Preferably, determining whether the parking space of the target parking space meets the preset parking standards specifically includes:

[0022] When the target parking space is perpendicular or angled relative to the direction of travel of the vehicle, the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is compared with the sum of the width of the vehicle plus the preset buffer distance. If the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is greater than the sum of the width of the vehicle plus the preset buffer distance, it is determined that the preset parking standard is met.

[0023] When the target parking space is a horizontal parking space relative to the direction of travel of the vehicle, the projected distance between the two closest points of the adjacent vehicles to the driving side in the direction of travel of the vehicle is compared with the sum of the length of the vehicle plus a preset buffer distance. If the projected distance between the two closest points of the adjacent vehicles to the driving side in the direction of travel of the vehicle is greater than the sum of the length of the vehicle plus the preset buffer distance, then the preset parking standard is met.

[0024] Preferably, the step of removing wall point cloud data from the above-ground point cloud data based on the available parking space specifically includes:

[0025] Identify the point cloud data that fits the wall surface from the point cloud data above the ground, and randomly sample multiple point clouds from it;

[0026] The sampled point cloud is fitted using a pre-defined planar model to obtain the fitting result;

[0027] Calculate the distance from other point clouds to the fitting result. If the distance from a certain point cloud to the fitting result is less than a preset distance threshold, then the point cloud is determined to be an interior point, and the number of all interior points is counted.

[0028] Repeat the process multiple times according to the preset number of iterations, and select the fitting result with the most interior points;

[0029] Calculate the ratio of the number of interior points to the number of points in each fitting result, and compare the ratio with a preset filtering threshold. Filter out ratios that are greater than the preset filtering threshold.

[0030] Preferably, identifying suspended obstacles within the available parking space specifically includes:

[0031] The system detects whether point cloud information exists in the vertical direction of the available parking space. If point cloud information exists, it determines that the existing point cloud information is a suspended obstacle attached to the wall, performs clustering processing on it, and outputs the suspended obstacle.

[0032] Preferably, it further includes:

[0033] If no point cloud information exists, the highest height information that the radar can reach is directly output, and the highest height information that the radar can reach is used as the height of the available parking space and combined with the available parking space space to output the final space parking space.

[0034] On the other hand, a spatial parking space recognition system is also provided to implement the aforementioned spatial parking space recognition method, including:

[0035] The point cloud data acquisition module is used to acquire point cloud data of the vehicle's surrounding environment through vehicle radar, and classify the point cloud data of the vehicle's surrounding environment to obtain ground point cloud data and above-ground point cloud data.

[0036] The obstacle recognition module is used to cluster the ground point cloud data to identify ground obstacles and determine available parking spaces based on the identified ground obstacles; based on the available parking spaces, it removes wall point cloud data from the above-ground point cloud data, and performs clustering on the removal results to identify suspended obstacles within the available parking spaces.

[0037] The space parking space calculation module is used to determine the minimum height value of the suspended obstacles based on the available parking space, and combine the minimum height value of the suspended obstacles as the height of the available parking space with the available parking space to output the final space parking space.

[0038] On the other hand, a car is also provided that identifies the vehicle's parking space using the aforementioned space parking space recognition system.

[0039] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0040] The spatial parking space recognition method, system, and vehicle provided by this invention accurately identify suspended obstacles (especially those attached to walls) in scenarios where parking spaces are scarce in underground parking lots. It outputs effective three-dimensional parking space information, and the output parking space height information provides the driving system or driver with crucial height information to determine whether a parking space is truly available in a parking scenario. This significantly improves parking space utilization while mitigating misjudgments caused by two-dimensional parking space output. Ultimately, it achieves a wide-range retrieval of available three-dimensional parking spaces and outputs high-precision information on available parking spaces. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention.

[0042] Figure 1 This is a schematic diagram of the main process of a spatial parking space recognition method in an embodiment of the present invention.

[0043] Figure 2 This is a schematic diagram of a parking space recognition system according to an embodiment of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 The diagram shown is a schematic representation of an embodiment of a parking space identification method provided by the present invention. In this embodiment, the method includes the following steps:

[0046] The vehicle acquires point cloud data of the surrounding environment using onboard radar, and classifies this data to obtain ground point cloud data and above-ground point cloud data. In other words, the vehicle detects parking spaces using onboard radar. In this embodiment, onboard radar refers to lidar. Since this is applied to a parking scenario, the point cloud information received from the lidar only needs to be captured within a 0 to 5 meter range, thus reasonably reducing the amount of data processing. Point cloud data refers to a set of vectors in a three-dimensional coordinate system. Scanned data is recorded in the form of points, each containing three-dimensional coordinates, and some may contain color information (RGB) or reflectance intensity information.

[0047] In this embodiment, ground points in the point cloud data are identified using preset point recognition rules, and the plane containing these ground points is designated as the ground plane. The ground plane is then calibrated using a preset transformation matrix to obtain ground point cloud data. Point cloud data excluding the ground point cloud data is output as above-ground point cloud data. This step essentially achieves ground segmentation, that is, identifying ground and non-ground elements through point cloud segmentation. The preset point recognition rules can refer to methods such as grid height difference, absolute height, and normal vectors for segmentation, aiming to find most ground points, locate the plane containing the points, and calibrate the plane using a transformation matrix, thereby segmenting the ground point cloud and reducing the impact on obstacle clustering.

[0048] Furthermore, the ground point cloud data is clustered to identify ground obstacles, and available parking spaces are determined based on the identified ground obstacles; that is, firstly, the available parking spaces on the ground are determined based on the situation of ground obstacles, and then, through the subsequent identification and judgment of suspended obstacles, the longitudinal (vertical height) parking space is determined; among them, the determination of available parking spaces is mainly based on whether the distance between two adjacent vehicles in the space meets the parking conditions.

[0049] In this embodiment, the point cloud information of target obstacles in the ground point cloud data is identified, and the point cloud information of the target obstacles is segmented according to a preset standard to obtain segmentation results. Each segment in the segmentation results is then clustered according to a preset minimum and maximum number of clustered point clusters, and the corresponding clustering results are output as ground obstacles. It is understood that, in terms of clustering, the preset clustering rules mentioned above can use common methods such as Euclidean clustering, which will not be elaborated here. Combined with segmented clustering, different clustering KD-tree radius thresholds are applied to the laser point cloud data obtained in regions with different distance ranges, while simultaneously setting the minimum and maximum number of clustered point clusters to complete the clustering of the target object point cloud and obtain the corresponding 3D Boundingbox results after clustering.

[0050] Specifically, the system determines whether the parking space of the target parking space meets the preset parking standards based on the range of ground obstacles. If the parking space meets the preset parking standards, the parking space is determined to be an available parking space; if the parking space does not meet the preset parking standards, the parking space is determined to be an unavailable parking space. When the lidar outputs available parking space, assuming the vehicle's driving direction is x, the bounding box of the target obstacle detected by the lidar can be used to further determine whether the unobstructed space meets the parking conditions. However, in actual parking, since the position of the parking space relative to the vehicle's driving direction (x direction) is not singular, there are multiple different situations, as described below.

[0051] The parking space of the target parking space meets the preset parking standard. Specifically, when the target parking space is perpendicular or angled relative to the direction of travel of the vehicle, the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is compared with the sum of the width of the vehicle and the preset buffer distance. If the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is greater than the sum of the width of the vehicle and the preset buffer distance, then the preset parking standard is met. It can be understood that in the case of perpendicular and angled parking spaces, when the projected distance on the x-axis between the two closest points of the two adjacent vehicles in the area bounded by the driving side is greater than the width of the vehicle plus the buffer distance, the parking condition is met.

[0052] When the target parking space is horizontal relative to the vehicle's driving direction, the projected distance between the two closest points of adjacent vehicles to the driving side in the vehicle's driving direction is compared with the sum of the vehicle's length and a preset buffer distance. If the projected distance between the two closest points of adjacent vehicles to the driving side in the vehicle's driving direction is greater than the sum of the vehicle's length and the preset buffer distance, then the preset parking standard is met. In the case of a horizontal parking space, the parking condition is met when the projected distance on the x-axis between the two closest points of adjacent vehicles within the bounded area near the driving side is greater than the vehicle length plus the buffer distance.

[0053] Furthermore, based on the available parking space, the wall point cloud data in the above-ground point cloud data is removed, and the removal results are clustered to identify suspended obstacles within the available parking space. That is, because a buffer distance is set when outputting the available parking space, the edges of the parking space may be flush with the wall. However, suspended obstacles on the wall cannot be directly clustered, resulting in the output 3D boundingbox being the smallest rectangular bounding box that includes both the wall and the suspended obstacle, potentially including spaces below the suspended obstacle that do not actually exist, leading to misjudgments. To improve this, the wall needs to be filtered out before clustering the obstacles flush with the wall to output a 3D boundingbox. The clustering method here is the same as the preset clustering rules (i.e., the clustering method for ground obstacles) mentioned above, and will not be elaborated further here.

[0054] In this embodiment, point cloud data adhering to the wall surface is identified from the point cloud data above the ground, and multiple point clouds are randomly sampled from them. The sampled point clouds are fitted using a preset planar model to obtain a fitting result. The distance from other point clouds to the fitting result is calculated. If the distance from a point cloud to the fitting result is less than a preset distance threshold, the point cloud is determined to be an interior point, and the number of all interior points is counted. This process is repeated multiple times according to a preset number of iterations, and the fitting result with the most interior points is selected. The ratio of the number of interior points in each fitting result to the total number of point clouds in the fitting result is calculated, and the ratio is compared with a preset filtering threshold. The ratios greater than the preset filtering threshold are filtered out. In essence, a planar model is fitted to K randomly sampled point clouds, the distance from other points to the fitted model is calculated, and interior points are identified. This process is repeated M times, and the planar fitting model with the most interior points is selected. After fitting all point clouds on the wall surface using the above planar model, the ratio of the number of interior points in each planar model to the total number of point clouds is calculated as m. A filtering threshold of M is set, and point clouds are filtered out from the plane containing the largest m exceeding the filtering threshold M. It should be noted that repeating the above steps a certain number of times (the specific number can be set according to the actual situation) can achieve the filtering of point clouds on walls (this also applies to scenes with multiple overlapping walls).

[0055] After the above filtering, suspended obstacles above the parking space can be accurately identified. The specific process is as follows: The system detects whether point cloud information exists in the vertical direction of the available parking space. If point cloud information exists, it is determined to be a suspended obstacle attached to the wall, and clustering is performed to output the suspended obstacle. Understandably, the system first detects whether point cloud information exists in the vertical direction of the parking space. If not, it directly outputs the maximum height achievable by the lidar radiation. If point cloud information exists, it can be determined to be a suspended obstacle attached to the wall after filtering out the wall. Since the wall has been filtered out, a clustering algorithm can be directly performed to output the actual 3D boundingbox of the suspended obstacle.

[0056] It should be noted that during the filtering process, if there are sporadic point clouds in the vertical space during detection (i.e., the detected point clouds cannot meet the clustering requirements and cannot reach the preset threshold), these sporadic power sources are determined to be interference and interference filtering is required.

[0057] Furthermore, based on the suspended obstacles within the available parking space, the minimum height value of the suspended obstacles is determined, and this minimum height value is used as the height of the available parking space. This is then combined with the available parking space to output the final parking space. In other words, by determining the two-dimensional space of the available parking space and combining it with the three-dimensional parking space composed of the vertically suspended space, effective three-dimensional parking space information is output.

[0058] In this embodiment, the system detects whether point cloud information exists in the vertical direction of the available parking space. If no point cloud information exists, the system directly outputs the highest height information achievable by radar radiation. This highest height information is then combined with the available parking space to output the final parking space. It is understood that if a floating obstacle is identified and its minimum height can be determined in the previous step of identifying floating obstacles, the vertical height value for parking is determined based on the minimum height of the floating obstacle. However, if no floating obstacle is identified, it means there are no floating obstacles within the highest detectable height of the vehicle-mounted radar (LiDAR). In this case, the highest detectable height of the LiDAR is used as the vertical height value for parking to determine the final parking space.

[0059] like Figure 2 As shown, the present invention also provides a parking space recognition system. In this embodiment, the system includes:

[0060] The point cloud data acquisition module is used to acquire point cloud data of the vehicle's surrounding environment through vehicle radar, and classify the point cloud data of the vehicle's surrounding environment to obtain ground point cloud data and above-ground point cloud data.

[0061] The obstacle recognition module is used to cluster the ground point cloud data to identify ground obstacles and determine available parking spaces based on the identified ground obstacles; based on the available parking spaces, it removes wall point cloud data from the above-ground point cloud data, and performs clustering on the removal results to identify suspended obstacles within the available parking spaces.

[0062] The space parking space calculation module is used to determine the minimum height value of the suspended obstacles based on the available parking space, and combine the minimum height value of the suspended obstacles as the height of the available parking space with the available parking space to output the final space parking space.

[0063] The present invention also provides a vehicle that uses the aforementioned space parking space recognition system to identify the vehicle's parking space.

[0064] It should be noted that the system described in the above embodiments corresponds to the method described in the above embodiments. Therefore, the parts of the system described in the above embodiments that are not described in detail can be obtained by referring to the content of the method described in the above embodiments, and will not be repeated here.

[0065] In summary, implementing the embodiments of the present invention has the following beneficial effects:

[0066] The spatial parking space recognition method, system, and vehicle provided by this invention accurately identify suspended obstacles (especially those attached to walls) in scenarios where parking spaces are scarce in underground parking lots. It outputs effective three-dimensional parking space information, and the output parking space height information provides the driving system or driver with crucial height information to determine whether a parking space is truly available in a parking scenario. This significantly improves parking space utilization while mitigating misjudgments caused by two-dimensional parking space output. Ultimately, it achieves a wide-range retrieval of available three-dimensional parking spaces and outputs high-precision information on available parking spaces.

[0067] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for identifying parking spaces, characterized in that, include: The vehicle acquires point cloud data of the surrounding environment through vehicle-mounted radar, and classifies the point cloud data of the surrounding environment to obtain ground point cloud data and above-ground point cloud data. Clustering is performed on the ground point cloud data to identify ground obstacles, and available parking spaces are determined based on the identified ground obstacles. Based on the available parking space, the wall point cloud data in the above-ground point cloud data is removed, and the removal results are clustered to identify suspended obstacles in the available parking space. The minimum height of the suspended obstacle is determined based on the suspended obstacle in the available parking space, and the minimum height of the suspended obstacle is used as the height of the available parking space and combined with the available parking space to output the final space parking space; Specifically, the step of removing wall point cloud data from the above-ground point cloud data based on the available parking space includes: Identify the point cloud data that fits the wall surface from the point cloud data above the ground, and randomly sample multiple point clouds from it; The sampled point cloud is fitted using a pre-defined planar model to obtain the fitting result; Calculate the distance from other point clouds to the fitting result. If the distance from a certain point cloud to the fitting result is less than a preset distance threshold, then the point cloud is determined to be an interior point, and the number of all interior points is counted. Repeat the process multiple times according to the preset number of iterations, and select the fitting result with the most interior points; Calculate the ratio of the number of interior points to the number of points in each fitting result, and compare the ratio with a preset filtering threshold. Filter out ratios that are greater than the preset filtering threshold.

2. The method as described in claim 1, characterized in that, The classification based on the point cloud data of the vehicle's surrounding environment yields ground point cloud data and above-ground point cloud data, specifically including: Ground points in the point cloud data are identified by a preset point recognition rule, and the plane where the ground points are located is taken as the ground plane; The ground plane is calibrated using a preset transformation matrix to obtain ground point cloud data; The point cloud data excluding the ground point cloud data is output as above-ground point cloud data.

3. The method as described in claim 2, characterized in that, The process of clustering the ground point cloud data to identify ground obstacles specifically includes: Identify the point cloud information of the target obstacle in the ground point cloud data, and segment the point cloud information of the target obstacle according to a preset standard to obtain the segmentation result; Cluster each segment in the segmentation result according to the preset minimum and maximum number of cluster points, and output the corresponding clustering results as ground obstacles.

4. The method as described in claim 1, characterized in that, The step of determining available parking space based on identified ground obstacles specifically includes: Determine whether the parking space of the target parking space meets the preset parking standards based on the extent of ground obstacles; If the parking space of the target parking space meets the preset parking standards, then the parking space is determined to be a usable parking space. If the parking space of the target parking space does not meet the preset parking standards, the parking space is determined to be an unusable parking space.

5. The method as described in claim 4, characterized in that, The determination of whether the parking space of the target parking space meets the preset parking standards specifically includes: When the target parking space is perpendicular or angled relative to the direction of travel of the vehicle, the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is compared with the sum of the width of the vehicle plus the preset buffer distance. If the projected distance between the two closest points of the two adjacent vehicles to the driving side in the direction of travel of the vehicle is greater than the sum of the width of the vehicle plus the preset buffer distance, it is determined that the preset parking standard is met. When the target parking space is a horizontal parking space relative to the direction of travel of the vehicle, the projected distance between the two closest points of the adjacent vehicles to the driving side in the direction of travel of the vehicle is compared with the sum of the length of the vehicle plus a preset buffer distance. If the projected distance between the two closest points of the adjacent vehicles to the driving side in the direction of travel of the vehicle is greater than the sum of the length of the vehicle plus the preset buffer distance, then the preset parking standard is met.

6. The method as described in claim 1, characterized in that, The identification of suspended obstacles within the available parking space specifically includes: The system detects whether point cloud information exists in the vertical direction of the available parking space. If point cloud information exists, it determines that the existing point cloud information is a suspended obstacle attached to the wall, performs clustering processing on it, and outputs the suspended obstacle.

7. The method as described in claim 6, characterized in that, Also includes: If no point cloud information exists, the highest height information that the radar can reach is directly output, and the highest height information that the radar can reach is used as the height of the available parking space and combined with the available parking space space to output the final space parking space.

8. A spatial parking space recognition system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The point cloud data acquisition module is used to acquire point cloud data of the vehicle's surrounding environment through vehicle radar, and classify the point cloud data of the vehicle's surrounding environment to obtain ground point cloud data and above-ground point cloud data. An obstacle recognition module is used to perform clustering processing on the ground point cloud data to identify ground obstacles and determine available parking spaces based on the identified ground obstacles. Based on the available parking space, wall point cloud data in the above-ground point cloud data is removed, and the removal results are clustered to identify suspended obstacles within the available parking space. The space parking space calculation module is used to determine the minimum height value of the suspended obstacles based on the available parking space, and combine the minimum height value of the suspended obstacles as the height of the available parking space with the available parking space to output the final space parking space.

9. A car, characterized in that, The spatial parking space recognition system as described in claim 8 is used to identify the vehicle's spatial parking space.

Images