A method, device and readable storage medium for identifying a ground marking

Abstract

Embodiments of the present application provide a ground mark recognition method and device, and a readable storage medium, wherein the method comprises: determining ground point cloud data; determining a point cloud skeleton from the ground point cloud data; the reflectivity of point data in the point cloud skeleton is higher than a reflectivity threshold; taking point cloud data around a position of the point cloud skeleton in the ground point cloud data as local ground point cloud data, determining foreground point cloud data from the local ground point cloud data; and recognizing geometric features of a ground mark based on the foreground point cloud data. Embodiments of the present application can accurately recognize geometric features of a ground mark.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of map data, and particularly relate to a ground marking recognition method and device and readable storage medium. BACKGROUND

[0002] Point cloud data is a collection of point data expressing the spatial distribution and surface characteristics of a target. Currently, a laser sensor on a high-precision acquisition vehicle emits laser to a certain point in space and measures the reflected light intensity (also known as reflectivity) and the time of reflection to obtain spatial position information, material information, etc. of the point. In combination with inertial navigation, position positioning and other surveying and mapping equipment, the spatial coordinates of the point are measured, so that the point data of the point is composed of the spatial coordinates, reflectivity, time of reflection, etc. of the point, and the point cloud data is formed by a large number of point data.

[0003] Currently, when producing high-precision map data or updating high-precision map data, it is necessary to recognize the geometric features of ground markings from the point cloud data. Therefore, how to provide a ground marking recognition scheme to accurately recognize the geometric features of ground markings has become a problem to be solved by those skilled in the art. SUMMARY

[0004] Therefore, embodiments of the present application provide a ground marking recognition method, device and readable storage medium to accurately recognize the geometric features of ground markings.

[0005] To achieve the above object, embodiments of the present application provide the following technical solutions:

[0006] A ground marking recognition method, comprising:

[0007] determining ground point cloud data;

[0008] determining a point cloud skeleton from the ground point cloud data; the reflectivity of point data in the point cloud skeleton is higher than a reflectivity threshold;

[0009] taking point cloud data around a position where the point cloud skeleton is located in the ground point cloud data as local ground point cloud data, and determining foreground point cloud data from the local ground point cloud data;

[0010] recognizing the geometric features of ground markings based on the foreground point cloud data.

[0011] A ground marking recognition device, comprising:

[0012] a ground point cloud determination module configured to determine ground point cloud data;

[0013] A point cloud skeleton determination module is configured to determine a point cloud skeleton from the ground point cloud data, wherein the reflectivity of point data in the point cloud skeleton is higher than a reflectivity threshold;

[0014] A foreground point cloud data determination module is configured to determine foreground point cloud data from local ground point cloud data, wherein the local ground point cloud data is point cloud data around a position of the point cloud skeleton in the ground point cloud data;

[0015] A ground marker feature recognition module is configured to recognize geometric features of a ground marker based on the foreground point cloud data.

[0016] A readable storage medium, wherein the readable storage medium stores a program for executing the ground marker recognition method.

[0017] The embodiment of the present application can determine a point cloud skeleton representing the shape of a ground element in the ground point cloud data after determining the ground point cloud data. Since the ground marker is generally a ground element of the foreground part of the ground, the embodiment of the present application can further determine foreground point cloud data from local ground point cloud data, wherein the local ground point cloud data is point cloud data around a position of the point cloud skeleton in the ground point cloud data, to realize the determination of point cloud data related to the ground marker. Furthermore, geometric features of the ground marker can be recognized based on the foreground point cloud data. The embodiment of the present application can recognize the geometric features of the ground marker from the foreground part of the ground point cloud data, avoid the information loss problem caused by converting the point cloud data into a raster image to recognize the geometric features of the ground marker, and accurately recognize the geometric features of the ground marker. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without any creative effort.

[0019] Figure 1 The flowchart of the ground marker recognition method provided by the embodiment of the present application;

[0020] Figure 2 The flowchart of the ground point cloud data determination method provided by the embodiment of the present application;

[0021] Figure 3 The example diagram of the point data connected region;

[0022] Figure 4 The flowchart of the point cloud skeleton determination method from the ground point cloud data provided by the embodiment of the present application;

[0023] Figure 5 A flowchart for determining the foreground point cloud data is provided for an embodiment of the present application;

[0024] Figure 6 A flowchart for determining the foreground point cloud data according to the first foreground point data connected region is provided for an embodiment of the present application;

[0025] Figure 7 A flowchart for extracting the point cloud data corresponding to the ground marking is provided for an embodiment of the present application;

[0026] Figure 8 A block diagram of the recognition device of the ground marking is provided for an embodiment of the present application;

[0027] Figure 9 A hardware block diagram of the computing device. DETAILED DESCRIPTION

[0028] It is an important work to determine the geometric features of the ground marking in the high-precision map data from the point cloud data for making or updating the high-precision map data.

[0029] Taking the updating of the high-precision map data as an example, the high-precision map data is mainly made by the point cloud data collected by the high-precision collection vehicle (such as RIEGL vehicle) on the road. Since the cost of the high-precision collection vehicle is high and the data collection period is long, if the point cloud data collected by the high-precision collection vehicle is used to update the high-precision map data when the high-precision map data is updated, there is no doubt that the cost of the update is high and the period is long. Therefore, at present, the three-dimensional reconstruction technology is generally used to update the local of the high-precision map data by using the road image data collected by the standard-precision collection vehicle (such as ADAS vehicle), so as to reduce the cost and period of updating the high-precision map data.

[0030] In the process of updating the high-precision map data by using the road image data collected by the standard-precision collection vehicle, one of the links is to align the ground elements of the point cloud data and the road image data. At this time, the geometric features of the ground elements need to be determined from the point cloud data and the road image data respectively, so as to realize the matching between the ground elements in the point cloud data and the road image data by matching the geometric features. As a typical ground element, the geometric features of the ground elements determined from the point cloud data involve the geometric features of the ground marking identified from the point cloud data, such as the ground marking of the lane line, the ground arrow and the marking indicating the traffic information. Therefore, when the high-precision map data is updated, it is an important work to process the point cloud data to identify the geometric features of the ground marking from the point cloud data.

[0031] A current way of identifying geometric features of ground markings from point cloud data is to project a ground portion in the point cloud data onto a horizontal plane to generate a raster image, and perform segmentation, identification and other processing on the raster image to identify geometric features of ground markings from the raster image. However, the point cloud data is obtained by emitting laser light from a laser sensor on a high-precision acquisition vehicle to a certain point in space. Since the interval between adjacent laser frames is large (for example, the interval between adjacent laser frames of a RIEGL vehicle is 5-10 cm), if the point cloud data is converted into a high-resolution raster image, there will be a large number of blanks between the effective pixels (at least one point data falls within the raster in the effective pixel), which affects the accuracy of the identified geometric features of the ground markings. Although the resolution of the raster image can be reduced to reduce the blanks between the effective pixels, the overall geometric features of the ground markings cannot be accurately identified from the raster image with a lower resolution. For example, the center line in the ground markings can be better identified from the raster image with a lower resolution, but the side line in the ground markings cannot be accurately identified.

[0032] It can be seen that the accuracy of identifying geometric features of ground markings in the prior art is low. Based on this, the embodiments of the present application provide a ground marking identification method, device and readable storage medium to accurately identify geometric features of ground markings from point cloud data. The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0033] As an optional implementation of the disclosure of the embodiments of the present application, Figure 1 An optional flow of a ground marking identification method provided by the embodiments of the present application is shown, which can be executed by a computing device with data processing capability. The computing device can be implemented by a server (for example, a single server or a server group composed of multiple servers), or by a terminal;

[0034] Referring to Figure 1 The ground marking identification method provided by the embodiments of the present application can include:

[0035] Step S10, determining ground point cloud data.

[0036] The ground mark is a ground element of a ground part in high-precision map data, and the embodiment of the present application can determine the ground point cloud data corresponding to the ground part from the collected point cloud data, thereby providing a basis for identifying the geometric features of the ground mark. The collected point cloud data can be input data of the embodiment of the present application, and is point cloud data collected by a high-precision collection vehicle on a road.

[0037] In an optional implementation, since there is no connection relationship between the spatial positions of the point data in the point cloud data, in order to describe the relationship of the point data with close spatial positions in the point cloud data, the concept of point data connected region is proposed here, that is, the point data with close spatial positions in the point cloud data can be considered to be connected, and the point data with close spatial positions can form a point data connected region, so there can be at least one point data connected region in which the point data is connected in the collected point cloud data. Optionally, the embodiment of the present application can consider the point data with a distance between the point data in the point cloud data not greater than a set distance threshold as connected point data by setting the distance threshold, thereby collecting the connected point data to form a point data connected region.

[0038] In an optional implementation, the embodiment of the present application can determine the point data connected region in which the point data is connected in the collected point cloud data, thereby determining the point data connected region corresponding to the ground part from the determined point data connected region, and the point data connected region corresponding to the ground part can be used as the ground point cloud data.

[0039] Step S11, determining a point cloud skeleton from the ground point cloud data; the reflectivity of the point data in the point cloud skeleton is higher than a reflectivity threshold.

[0040] The point cloud skeleton can represent the shape of the ground element in the ground point cloud data; the embodiment of the present application can determine the point cloud skeleton from the ground point cloud data by a point cloud skeleton extraction method, and limit the reflectivity of the point data in the point cloud skeleton to be higher than the reflectivity threshold; the embodiment of the present application is not limited to the point cloud skeleton extraction method used, and as an optional implementation, the embodiment of the present application can perform median filtering processing and Laplace filtering processing on the point data in the ground point cloud data, then remove the point data with a reflectivity lower than the reflectivity threshold, and then determine the point data connected region by determining the point data connected region, and determine the point cloud skeleton from the determined point data connected region.

[0041] The point cloud skeleton can be used to understand the geometric features of an object, and the embodiment of the present application determines the point cloud skeleton from the ground point cloud data, and can extract the point cloud data representing the shape of the ground element from the ground point cloud data.

[0042] Step S12, taking the point cloud data in the periphery of the position where the point cloud skeleton is located in the ground point cloud data as local ground point cloud data, determining foreground point cloud data from the local ground point cloud data.

[0043] After the point cloud skeleton is determined from the ground point cloud data, since the ground element contains other elements in addition to the ground mark, and the ground mark is generally the ground element of the foreground part of the ground, the embodiment of the application can take the point cloud data in the periphery of the position where the point cloud skeleton is located in the ground point cloud data as local ground point cloud data, thereby determining the foreground point cloud data of the foreground part in the local ground point cloud data, and the determined foreground point cloud data can contain the point cloud data corresponding to the ground mark.

[0044] Step S13, identifying the geometric features of the ground mark based on the foreground point cloud data.

[0045] After the foreground point cloud data is obtained, the foreground point cloud data contains the point cloud data corresponding to the ground mark, and therefore the embodiment of the application can identify the geometric features of the ground mark based on the foreground point cloud data.

[0046] Optionally, the embodiment of the application can extract the point cloud data corresponding to the ground mark from the foreground point cloud data, and thereby identify the geometric features of the ground mark according to the point cloud data corresponding to the ground mark.

[0047] In a more specific optional implementation, the embodiment of the application can set a more fine distance threshold, thereby determining the point data connected region of the point data connected in the foreground point cloud data according to the more fine distance threshold, and further removing the point data connected region whose number of point data is lower than the number threshold, to filter the points connected loosely in the foreground point cloud data, and the remaining set of point data can be used as the point cloud data corresponding to the ground mark. After the point cloud data corresponding to the ground mark is obtained, the embodiment of the application can perform a concave hull processing on the point cloud data corresponding to the ground mark, to identify the geometric features of the ground mark.

[0048] The identification method of the ground mark provided by the embodiment of the application can determine the ground point cloud data, determine the point cloud skeleton from the ground point cloud data, the reflectivity of the point data in the point cloud skeleton being higher than a reflectivity threshold, take the point cloud data in the periphery of the position where the point cloud skeleton is located in the ground point cloud data as local ground point cloud data, determine foreground point cloud data from the local ground point cloud data, and further identify the geometric features of the ground mark based on the foreground point cloud data.

[0049] It can be seen that, after the ground point cloud data is determined, the point cloud skeleton representing the shape of the ground elements in the ground point cloud data is determined, and since the ground marker is generally the ground element of the foreground part of the ground, the embodiment of the present application can further take the point cloud data around the position of the point cloud skeleton in the ground point cloud data as the local ground point cloud data, determine the foreground point cloud data from the local ground point cloud data, and realize the determination of the point cloud data related to the ground marker. Furthermore, based on the foreground point cloud data, the geometric features of the ground marker are identified. The embodiment of the present application can identify the geometric features of the ground marker from the foreground part of the ground point cloud data, avoid the information loss problem caused by converting the point cloud data into a raster image to identify the geometric features of the ground marker, and accurately identify the geometric features of the ground marker.

[0050] As an optional implementation of the disclosure of the embodiment of the present application, the embodiment of the present application can acquire the collected point cloud data, and determine the ground point cloud data corresponding to the ground part from the collected point cloud data. In an optional specific implementation, Figure 2 An optional flow for determining the ground point cloud data provided by the embodiment of the present application is shown, referring to Figure 2 The flow can include:

[0051] Step S20, determining a point data connected region of the point data connected to each other from the collected point cloud data.

[0052] The point data connected region refers to a collection of point data with close spatial positions in the point cloud data. Optionally, the embodiment of the present application can use the distance between the point data to represent the closeness of the spatial positions between the point data. By setting a distance threshold, the embodiment of the present application can determine a collection of point data with a distance between the point data in the point cloud data not greater than the set distance threshold as a point data connected region. Furthermore, the embodiment of the present application can determine the distance between the point data in the collected point cloud data, and collect the point data with a distance between the point data in the collected point cloud data not greater than the set distance threshold to form a point data connected region.

[0053] For example, as Figure 3 shown in the example of the point data connected region, one black dot in the figure can represent one point data. For the collected point cloud data c, at least one point data connected region can be determined from c, wherein the i-th point data connected region s_i can be represented as: for any point data p in s_i, the distance between the point data p and another point data q different from the point data p in s_i is not greater than the set distance threshold.

[0054] Optionally, the distance between point data referred to in the embodiments of the present invention can refer to the two-dimensional Euclidean distance between point data. Optionally, the set distance threshold referred to in the embodiments of the present invention can be selected according to the nature of the collected point cloud data. For example, if the spacing between adjacent laser frames of a RIEGL vehicle is between 5 cm and 10 cm, then the set distance threshold can be optionally set to 0.1 m. Of course, the specific value of the set distance threshold here is only an exemplary illustration.

[0055] In another optional implementation, since the ground is often relatively flat, that is, the elevation changes between the corresponding point data are continuous, the embodiments of the present invention can combine the distance between the point data and the elevation change rate to determine the connected regions of the point data from the collected point cloud data; optionally, on the basis that the distance between the point data is not greater than a set distance threshold, the embodiments of the present invention can also add an elevation continuity constraint, so that when the elevation change rate between adjacent point data is less than the elevation change rate threshold, the two point data are considered to belong to the same connected region of point data;

[0056] Accordingly, embodiments of the present invention can determine the distance between each point data in the collected point cloud data, as well as the elevation change rate between adjacent point data, thereby classifying the set of point data whose elevation change rate between adjacent point data is less than the elevation change rate threshold and whose distance between point data is not greater than a set distance threshold as the same point data connected region, thereby determining the connected regions of each point data from the collected point cloud data.

[0057] Step S21: Among the determined point data connected regions, the point data connected region with the largest number of point data is determined as the ground point cloud data.

[0058] The inventors of this invention discovered through analysis that more than 50% of the point cloud data collected on roads belongs to the ground portion. Therefore, the connected regions of point data with the largest number of connected points in the point cloud data collected on roads can be considered as ground point cloud data. Based on this, after determining the connected regions of each point data in the point cloud data collected on roads, the embodiments of this invention can determine the number of point data in each connected region, thereby identifying the connected region of the point data with the largest number of point data as ground point cloud data.

[0059] pass Figure 2 The method shown in this embodiment of the invention can determine the ground point cloud data of the ground portion from the collected point cloud data, thereby providing a basis for subsequent identification of the geometric features of ground markers in the ground portion; of course, Figure 2 The method shown is only one possible way to determine ground point cloud data from the collected point cloud data. This embodiment of the invention can also support other possible ways to determine ground point cloud data.

[0060] As an optional implementation of the disclosure of the embodiments of the present application, Figure 4 An optional process for determining a point cloud skeleton from ground point cloud data is shown, which can be considered as Figure 1 As an optional implementation of step S11, refer to Figure 4 The process can include:

[0061] Step S30, performing median filtering processing on the reflectivity of the point data in the ground point cloud data to obtain first point cloud data.

[0062] To remove noise, the embodiments of the present application can perform median filtering processing on the point data in the ground point cloud data. The point cloud data after median filtering processing on the reflectivity of the point data can be referred to as first point cloud data.

[0063] Optionally, the embodiments of the present application can determine the intermediate reflectivity of the neighboring point data corresponding to the point data in the ground point cloud data, and determine the intermediate reflectivity as the reflectivity of the point data, thereby achieving median filtering processing on the reflectivity of the point data in the ground point cloud data.

[0064] Specifically, for any point data in the ground point cloud data, the embodiments of the present application can determine the neighboring point data of the point data from the ground point cloud data. Optionally, the embodiments of the present application can determine the neighboring point data within a set distance threshold from the point data from the ground point cloud data. The reflectivity of the determined neighboring point data is sorted, and the intermediate reflectivity corresponding to the sorting is determined as the reflectivity of the point data, thereby achieving median filtering processing on the reflectivity of the point data in the ground point cloud data.

[0065] Optionally, in the process of determining the intermediate reflectivity of the neighboring point data corresponding to the point data in the ground point cloud data, if the number of neighboring point data of any point data in the ground point cloud data is odd, the reflectivity of the neighboring point data is sorted, and the reflectivity of the intermediate position corresponding to the sorting is determined as the intermediate reflectivity. If the number of neighboring point data of any point data in the ground point cloud data is even, the reflectivity of the neighboring point data is sorted, and the average of the two reflectivities of the intermediate position corresponding to the sorting is determined as the intermediate reflectivity.

[0066] For example, for any point data p in the ground point cloud data, the neighboring point data within a set distance threshold from the point data p can be determined, and the reflectivity of the neighboring point data is sorted. If the number of neighboring point data is odd, the reflectivity of the point data in the middle of the sorting is determined as the reflectivity of the point data p. If the number of neighboring point data is even, the average of the reflectivity of the two point data in the middle of the sorting is determined as the reflectivity of the point data p.

[0067] Step S31, performing Laplace filtering on the reflectivity of the point data in the first point cloud data to obtain second point cloud data.

[0068] To measure the influence of the reflectivity of the neighboring point data around the point data in the first point cloud data on the reflectivity of the point data, the embodiment of the present application can perform Laplace filtering on the reflectivity of the point data in the first point cloud data, thereby optimizing the reflectivity of the point data to obtain second point cloud data. That is, the second point cloud data can be considered as the point cloud data formed after the reflectivity of the point data in the first point cloud data is re- assigned by performing Laplace filtering on the reflectivity of the point data in the first point cloud data.

[0069] Optionally, the optional implementation of performing Laplace filtering on the reflectivity of the point data in the first point cloud data can be: for any point data in the first point cloud data, determining the neighboring point data of the point data in the first point cloud data. In an optional implementation, the embodiment of the present application can determine the neighboring point data in the first point cloud data having a distance from the point data within a set distance threshold. Determining the reflectivity difference between the point data and each of the neighboring point data, thereby adding the determined reflectivity differences and dividing the number of the neighboring point data of the point data, and taking the absolute value of the calculation result as the reflectivity of the point data.

[0070] For example, for any point data p in the first point cloud data, the embodiment of the present application can determine each neighboring point data q having a distance from the point data p within a set distance threshold, calculate the reflectivity difference between the point data p and each of the neighboring point data q, add the reflectivity differences and divide the number of the neighboring point data q, and take the absolute value of the calculation result as the reflectivity of the point data p. At this time, the reflectivity of the point data p measures the degree of change of the reflectivity of the neighboring point data q around the point data p, that is, the reflectivity difference intensity of the neighboring point data q around the point data p.

[0071] Step S32, filtering the point data in the second point cloud data having a reflectivity lower than a set reflectivity threshold to obtain third point cloud data.

[0072] Optionally, the set reflectivity threshold can be obtained according to experience, and the embodiment of the present application is not limited to the specific value of the set reflectivity threshold. Generally speaking, the greater the set reflectivity threshold, the less sensitive the point data in the point cloud data to the change of the reflectivity, and the smaller the set reflectivity threshold, the more sensitive the point data in the point cloud data to the change of the reflectivity.

[0073] Step S33, determining the point data connected region of the point data in the third point cloud data.

[0074] Optionally, the introduction of step S33 can refer to the description of the corresponding point data connected region determined in the foregoing, which will not be described herein again. In an example, the present embodiment can set the point data set in which the distance between the point data in the third point cloud data is not greater than the set distance threshold as the same point data connected region.

[0075] Step S34, from the determined point data connected region, remove the point data connected region in which the number of point data is lower than the number threshold, to obtain the point cloud skeleton.

[0076] Optionally, the number threshold can be obtained according to experience, and the present embodiment is not limited to the specific value of the number threshold. Generally speaking, the greater the number threshold is, the more accurate the contour of the ground marking identified subsequently will be, but the recall rate will also be reduced. Therefore, the specific value of the number threshold can be determined on the basis of measuring the accuracy and recall rate of the contour of the ground marking.

[0077] It can be seen that the present embodiment can perform median filtering processing on the reflectivity of the point data in the ground point cloud data, and then perform Laplace filtering processing on the reflectivity of the point data after the median filtering processing. The point data in which the reflectivity is lower than the set reflectivity threshold is filtered after the Laplace filtering processing. For the remaining point data after the filtering, the present embodiment can determine the point data connected region in which the point data are connected, remove the point data connected region in which the number of point data is lower than the number threshold from the determined point data connected region, and realize the determination of the point cloud skeleton. The point cloud skeleton can be point cloud data representing the shape of the ground element in the ground point cloud data.

[0078] In optional implementations, if the number of point cloud skeletons is more than two, the present embodiment can also combine a plurality of point cloud skeletons in which the distance between the point cloud skeletons is close into one point cloud skeleton, so as to perform step S12 again, i.e., step S12 can be performed on the basis of the point cloud skeleton after the point cloud skeleton combination processing. Figure 1 Step S12, i.e., step S12 can be processed on the basis of the point cloud skeleton after the point cloud skeleton combination processing. In optional implementations, the present embodiment can set a predetermined combination distance, so as to combine the point cloud skeletons in which the distance between the point cloud skeletons is lower than the predetermined combination distance. After the point cloud skeleton combination processing, the number of point cloud skeletons can be at least one. Optionally, the distance between the point cloud skeletons can be represented by the distance between the center points of the point cloud skeletons. Of course, the combination processing of the point cloud skeletons is only an optional means.

[0079] After the determination of the point cloud skeleton, as one of the optional implementations of the disclosure of the present embodiment, Figure 5 An optional flow for determining the foreground point cloud data is shown, which is described with reference to Figure 5 The flow can include:

[0080] Step S40, collecting point data in a local distance around the position of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data.

[0081] Optionally, the embodiment of the present application can collect point data in a local distance around the position of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data. Specifically, the embodiment of the present application can collect point data in a local distance around the position of each point cloud skeleton in the ground point cloud data according to the position of each point cloud skeleton to obtain local ground point cloud data corresponding to the position around each point cloud skeleton.

[0082] For example, C1_i represents the i-th point cloud skeleton, and B represents ground point cloud data. The embodiment of the present application can collect point data in a local distance around the position of C1_i in the ground point cloud data B according to the position of C1_i to form local ground point cloud data B_i corresponding to the position around C1_i.

[0083] Step S41, determining first local foreground ground point cloud data corresponding to the foreground part from the local ground point cloud data, and determining a first foreground point data connected region in which point data is connected in communication from the first local foreground ground point cloud data, wherein the number of point data in the first foreground point data connected region is not less than a number threshold.

[0084] After determining the local ground point cloud data corresponding to the position around the point cloud skeleton in step S40, the embodiment of the present application can distinguish point cloud data of the foreground part from the local ground point cloud data. The point cloud data of the foreground part in the local ground point cloud data can be referred to as first local foreground ground point cloud data.

[0085] In an optional implementation, the foreground part and the background part in the local ground point cloud data can be distinguished by a foreground reflectivity threshold. The embodiment of the present application can determine a foreground reflectivity threshold of the local ground point cloud data, filter point data in the local ground point cloud data whose reflectivity is lower than the foreground reflectivity threshold to obtain point cloud data of the foreground part in the local ground point cloud data (i.e., first local foreground ground point cloud data).

[0086] Optionally, the method for calculating the foreground reflectance threshold value can be various, such as a histogram method, a clustering method, etc., and the embodiments of the present application are not limited thereto. For example, the Kittler's minimum error threshold method (MET) is taken as an example to illustrate the process of determining the foreground reflectance threshold value in the embodiments of the present application: a histogram of the local ground point cloud data is established, the energy value of the local ground point cloud data under various reflectance threshold value possibilities is calculated according to the histogram, and an energy function is obtained; a first reflectance value corresponding to the maximum value in the energy function is found, and a second reflectance value corresponding to the first peak value of the energy function from the right side is found, and the reflectance value corresponding to the valley value between the first reflectance value and the second reflectance value is determined as the foreground reflectance threshold value of the local ground point cloud data; in this process, if the first reflectance value is equal to the second reflectance value, the emissivity value corresponding to the maximum value between the two gradient valley values of the energy function to the right of the first reflectance value is found as the foreground reflectance threshold value; if the above methods fail, the average value of the first reflectance value and the maximum reflectance value is directly taken as the foreground reflectance threshold value;

[0087] Since the reflectance of the point data in the foreground part is higher, the embodiments of the present application can filter the point data in the local ground point cloud data whose reflectance is lower than the foreground reflectance threshold value, so as to obtain the point cloud data in the foreground part of the local ground point cloud data (i.e., the first local foreground ground point cloud data).

[0088] The embodiments of the present application can determine the foreground reflectance threshold value for the local ground point cloud data corresponding to the position periphery of each point cloud skeleton according to the above method, so as to determine the point cloud data in the foreground part of each local ground point cloud data based on the foreground reflectance threshold value determined for each local ground point cloud data.

[0089] It can be seen that, after the ground point cloud data is divided into different local ground point cloud data, the embodiments of the present application can respectively determine the corresponding foreground reflectance threshold value for different local ground point cloud data, and then divide the foreground part and the background part in the local ground point cloud data according to the corresponding foreground reflectance threshold value for different local ground point cloud data, so that different local ground point cloud data in the ground point cloud data uses different foreground reflectance threshold values to distinguish the foreground part and the background part. Compared with the division of the foreground part and the background part of the ground point cloud data using the same foreground reflectance threshold value, which cannot accurately achieve the division of the foreground part and the background part (for example, the reflectance of the ground corresponding to the high-precision acquisition vehicle close to the high-precision acquisition vehicle is higher than the reflectance of the ground far from the high-precision acquisition vehicle, the reflectance of the ground mark with low abrasion degree is higher than the reflectance of the ground mark with high abrasion degree, etc., so that the accuracy of the division result is not high when the ground point cloud data is divided into the foreground part and the background part using the same foreground reflectance threshold value), the embodiments of the present application can accurately divide the foreground part in the local ground point cloud data.

[0090] After obtaining the first local foreground ground point cloud data corresponding to the foreground part of the local ground point cloud data, the embodiment of the application can determine the point data connected region of the point data in the first local foreground ground point cloud data, so as to remove the point data connected region in which the number of point data is lower than the number threshold from the determined point data connected region, and obtain the first foreground point data connected region, so that the number of point data in the first foreground point data connected region is not lower than the number threshold. The first foreground point data connected region can be considered as the point data connected region of the foreground part of the ground point cloud data.

[0091] The introduction of determining the point data connected region can refer to the corresponding description of determining the point data connected region in the foregoing, which will not be described again this time; in an example, the embodiment of the application can regard the point data set in which the distance between the point data in the first local foreground ground point cloud data is not greater than the set distance threshold as the same point data connected region.

[0092] Step S42, determining the foreground point cloud data according to the first foreground point data connected region.

[0093] In an optional implementation, after obtaining the first foreground point data connected region, the embodiment of the application can directly regard the first foreground point data connected region as the foreground point cloud data determined from the local ground point cloud data.

[0094] In another optional implementation, after obtaining the first foreground point data connected region, the embodiment of the application can regard the first foreground point data connected region as the point cloud skeleton, re-determine the corresponding local ground point cloud data, and then determine the second foreground point data connected region in which the number of point data of the foreground part is not lower than the number threshold from the re-determined local ground point cloud data, and determine whether the first foreground point data connected region is reliable by comparing the number of point data of the first foreground point data connected region and the second foreground point data connected region, so as to determine whether to determine the first foreground point data connected region as the foreground point cloud data.

[0095] In an optional implementation, Figure 6 An optional process of determining the foreground point cloud data according to the first foreground point data connected region provided by the embodiment of the application is shown, which refers to Figure 6 The process can include:

[0096] Step S50, collecting the point data around the position of the first foreground point data connected region in the ground point cloud data.

[0097] The result of the collection can be considered as the corresponding local ground point cloud data of the first foreground point data connected region as the point cloud skeleton.

[0098] In step S51, the second local foreground ground point cloud data corresponding to the foreground part is determined from the collected point data, and a second foreground point data connected region in which the point data is connected is determined from the second local foreground ground point cloud data, wherein the number of point data of the second foreground point data connected region is not less than the number threshold.

[0099] The optional implementation of determining the second local foreground ground point cloud data corresponding to the foreground part from the collected point data, and the optional implementation of determining the second foreground point data connected region in which the point data is connected from the second local foreground ground point cloud data can be realized according to the foregoing description, and will not be described here.

[0100] In step S52, if the difference between the number of point data of the second foreground point data connected region and the first foreground point data connected region is within a predetermined error, the first foreground point data connected region is determined as the foreground point cloud data.

[0101] In step S53, if the difference between the number of point data of the second foreground point data connected region and the first foreground point data connected region is not within a predetermined error, the first foreground point data connected region is removed.

[0102] It can be seen that for any point cloud skeleton, in the first processing, the embodiment of the application can determine the foreground point data connected region corresponding to the foreground part from the ground point cloud data according to the reference point cloud with the point cloud skeleton as the reference point cloud; in the second processing, the embodiment of the application can determine the foreground point data connected region corresponding to the foreground part from the ground point cloud data according to the reference point cloud with the foreground point data connected region determined in the last time as the reference point cloud, so that when the difference between the number of point data of the foreground point data connected regions determined in adjacent two times is within a predetermined error, the foreground point data connected region determined in the first time is retained as the determined foreground point cloud data.

[0103] For example, taking C1_i as the i th point cloud skeleton, the embodiment of the application can first take C1_i as the reference point cloud, and determine the foreground point data connected region D_i corresponding to the foreground part from the ground point cloud data according to the reference point cloud; then, take D_i as the reference point cloud, and determine the foreground point data connected region D_i' corresponding to the foreground part from the ground point cloud data according to the reference point cloud; if the number of point data of D_i' and D_i differs greatly (for example, the difference between the number of point data of D_i' and D_i exceeds a predetermined error), it is considered that the reliability of D_i is low, and D_i can be removed; if the difference between the number of point data of D_i' and D_i is lower than the predetermined number error, it is considered that D_i is reliable, and D_i can be taken as the determined foreground point cloud data.

[0104] After determining the foreground point cloud data, the embodiment of the application can extract the point cloud data corresponding to the ground marker from the foreground point cloud data, and identify the geometric features of the ground marker according to the point cloud data corresponding to the ground marker, so as to realize the identification of the geometric features of the ground marker in the ground point cloud data. As an optional implementation of the disclosure of the embodiment of the application, Figure 7 An optional process for extracting point cloud data corresponding to the ground marker is shown, and the process can include Figure 7 The process can include

[0105] Step S60, determining the distance between the point data and the nearest neighbor point data in the foreground point cloud data.

[0106] For each foreground point cloud data, the embodiment of the application can calculate the distance between each point data and the nearest neighbor point data in the foreground point cloud data.

[0107] Step S61, determining the connected distance threshold corresponding to the foreground point cloud data according to the average value and the standard deviation of the distance.

[0108] After determining the distance between each point data and the nearest neighbor point data in the local ground foreground data, the embodiment of the application can determine the average value and the standard deviation of the distance according to the determined distances, and then determine the connected distance threshold corresponding to the local ground foreground data according to the average value and the standard deviation.

[0109] For example, the embodiment of the application can add the average value to 3 times the standard deviation to obtain the connected distance threshold, assuming that the average value of the distance is m, the standard deviation of the distance is d, and the connected distance threshold is s, then the connected distance threshold s can be expressed as m+3*d. The connected distance threshold can be considered as a more fine distance threshold set for the foreground point cloud data, and the embodiment of the application can determine a more fine point data connected region for the foreground point cloud data based on the connected distance threshold.

[0110] Step S62, according to the connectivity distance threshold, determine the point data connectivity region of the point data connectivity in the foreground point cloud data.

[0111] The embodiment of the application can set the connectivity distance threshold as a distance threshold, and perform point data connectivity region searching on the foreground point cloud data, so as to determine the point data connectivity region of the point data connectivity in the foreground point cloud data. Optionally, the embodiment of the application can determine the point data whose distance between points is not greater than the connectivity distance threshold in the foreground point cloud data, so as to collect the point data whose distance between points is not greater than the connectivity distance threshold in the connectivity distance threshold, and obtain the determined point data connectivity region.

[0112] Step S63, remove the point data connectivity region whose number of point data is lower than the number threshold in the determined point data connectivity region, and collect the remaining point data as the point cloud data corresponding to the ground marker.

[0113] After the point data connectivity region is determined from the foreground point cloud data according to the connectivity distance threshold, the embodiment of the application can remove the point data connectivity region whose number of point data is lower than the number threshold, so as to collect the remaining point data together as the point cloud data corresponding to the ground marker.

[0114] The ground marker identification method provided by the embodiment of the application can identify the point cloud data corresponding to the ground marker from the foreground part of the local ground according to the position and reflectivity information of the point cloud data, so as to realize the identification of the geometric feature of the ground marker, avoid the information loss problem of the point cloud data caused by the conversion of the point cloud data to the raster image, and realize the accurate geometric feature of the ground marker.

[0115] Further, when determining the ground point cloud data, the embodiment of the application can use the continuity of the height variation of the point data to constrain the determination of the ground point cloud data, so that the accuracy of the determined ground point cloud data is higher.

[0116] Further, the embodiment of the application uses different foreground reflectivity thresholds for different local ground point cloud data, distinguishes the foreground part and the background part, so that the foreground part (i.e. the foreground point cloud data) in the local ground point cloud data determined by the embodiment of the application is more accurate, and the accuracy of the geometric shape of the subsequently determined ground marker can be ensured.

[0117] Further, after determining the foreground point cloud data, the embodiment of the application can determine the point data connectivity region through a more fine connectivity distance threshold, so that the geometric shape of the subsequently determined ground marker has a higher recall rate.

[0118] The above describes a plurality of embodiment schemes provided by the embodiments of the present application, and each optional mode introduced by each embodiment scheme can be combined with each other and cross-referenced without conflict, thereby extending a plurality of possible embodiment schemes, which can be considered as the embodiment schemes disclosed and disclosed by the embodiments of the present application.

[0119] The following describes a ground mark recognition device provided by the embodiments of the present application. The ground mark recognition device described below can be considered as a program function module required to be set to implement the ground mark recognition method provided by the embodiments of the present application. The content of the ground mark recognition device described below can be mutually corresponding and referred to with the content of the ground mark recognition method described above.

[0120] As an optional implementation, Figure 8 A block diagram of the ground mark recognition device provided by the embodiments of the present application is shown, referring to Figure 8 The ground mark recognition device can include:

[0121] A ground point cloud determination module 100 is configured to determine ground point cloud data.

[0122] A point cloud skeleton determination module 200 is configured to determine a point cloud skeleton from the ground point cloud data; and the reflectivity of point data in the point cloud skeleton is higher than a reflectivity threshold.

[0123] A foreground point cloud data determination module 300 is configured to determine, as local ground point cloud data, point cloud data around a position of the point cloud skeleton in the ground point cloud data, and determine foreground point cloud data from the local ground point cloud data.

[0124] A ground mark feature recognition module 400 is configured to recognize geometric features of a ground mark based on the foreground point cloud data.

[0125] Optionally, the ground point cloud determination module 100 configured to determine ground point cloud data can specifically include:

[0126] Acquire the collected point cloud data.

[0127] Determine ground point cloud data corresponding to a ground portion from the collected point cloud data.

[0128] Optionally, the ground point cloud determination module 100 configured to determine ground point cloud data corresponding to a ground portion from the collected point cloud data can specifically include:

[0129] Determine a point data connected region in which point data are connected from the collected point cloud data.

[0130] Determine, as the ground point cloud data, a point data connected region in which the number of point data is the largest among the determined point data connected regions.

[0131] Optionally, the ground point cloud determining module 100, configured to determine point data connected regions of the point data connected to each other from the collected point cloud data, can specifically include:

[0132] determine the distance between each point data in the collected point cloud data and the elevation change rate between adjacent point data;

[0133] a point data set with an elevation change rate between adjacent point data less than an elevation change rate threshold and a distance between point data not greater than a set distance threshold is determined as a same point data connected region.

[0134] Optionally, the point cloud skeleton determining module 200, configured to determine a point cloud skeleton from the ground point cloud data, can specifically include:

[0135] perform median filtering processing on the reflectivity of the point data in the ground point cloud data;

[0136] perform Laplace filtering processing on the reflectivity of the point data after the median filtering processing;

[0137] filter the point data after the Laplace filtering processing with reflectivity lower than a set reflectivity threshold;

[0138] determine point data connected regions of the point data connected to each other in the filtered point data, remove point data connected regions with a number of point data lower than a number threshold from the determined point data connected regions, and obtain the point cloud skeleton.

[0139] Optionally, the point cloud skeleton determining module 200, configured to perform median filtering processing on the reflectivity of the point data in the ground point cloud data, can specifically include:

[0140] determine the intermediate reflectivity of the adjacent point data corresponding to the point data in the ground point cloud data as the reflectivity of the point data.

[0141] Optionally, the point cloud skeleton determining module 200, configured to determine the intermediate reflectivity of the adjacent point data corresponding to the point data in the ground point cloud data, can specifically include:

[0142] for any point data in the ground point cloud data, determine the adjacent point data of the point data from the ground point cloud data;

[0143] sort the reflectivity of the determined adjacent point data;

[0144] if the number of the determined adjacent point data is odd, determine the reflectivity at the middle position of the sorting as the intermediate reflectivity;

[0145] If the determined number of neighboring point data is even, the mean value of the two reflectivities ranked in the middle is determined as the intermediate reflectivity.

[0146] Optionally, the point cloud skeleton determining module 200 is configured to perform Laplace filtering on the reflectivity of the point data after the median filtering, and can specifically include:

[0147] For any point data after the median filtering, the neighboring point data of the point data is determined.

[0148] The reflectivity difference between the point data and each of the neighboring point data is determined, the determined reflectivity differences are added and divided by the number of the neighboring point data of the point data, and the absolute value of the calculation result is taken as the reflectivity of the point data.

[0149] Optionally, the ground mark recognition device provided by the embodiment of the present application can also be used to: if the point cloud skeleton determined by the point cloud skeleton determining module 200 is more than two, the point cloud skeletons with a distance lower than a predetermined merging distance can be merged.

[0150] Optionally, the foreground point cloud data determining module 300 is configured to take the point cloud data around the position of the point cloud skeleton in the ground point cloud data as local ground point cloud data, and determine the foreground point cloud data from the local ground point cloud data, and can specifically include:

[0151] The point data around the position of the point cloud skeleton in the ground point cloud data is collected to obtain local ground point cloud data.

[0152] The first local foreground ground point cloud data corresponding to the foreground part is determined from the local ground point cloud data, and the first foreground point data connected region in which the point data is connected is determined from the first local foreground ground point cloud data, wherein the number of point data in the first foreground point data connected region is not less than a number threshold.

[0153] The foreground point cloud data is determined according to the first foreground point data connected region.

[0154] Optionally, the foreground point cloud data determining module 300 is configured to determine the foreground point cloud data according to the first foreground point data connected region, and can specifically include:

[0155] The point data around the position of the first foreground point data connected region in the ground point cloud data is collected.

[0156] The second local foreground ground point cloud data corresponding to the foreground part is determined from the collected point data, and the second foreground point data connected region in which the point data is connected is determined from the second local foreground ground point cloud data, wherein the number of point data in the second foreground point data connected region is not less than a number threshold.

[0157] if the difference between the number of point data of the second foreground point data connected region and the first foreground point data connected region is within a predetermined error, determining the first foreground point data connected region as foreground point cloud data;

[0158] if the difference between the number of point data of the second foreground point data connected region and the first foreground point data connected region is not within a predetermined error, removing the first foreground point data connected region.

[0159] Optionally, the foreground point cloud data determining module 300 is configured to determine first local foreground ground point cloud data corresponding to a foreground part from the local ground point cloud data, and can specifically include:

[0160] determining a foreground reflectivity threshold of the local ground point cloud data;

[0161] filtering point data with reflectivity lower than the foreground reflectivity threshold in the local ground point cloud data to obtain first local foreground ground point cloud data.

[0162] Optionally, the foreground point cloud data determining module 300 is configured to determine a first foreground point data connected region in which point data are connected from the first local foreground ground point cloud data, and can specifically include:

[0163] determining a point data connected region in which point data are connected in the first local foreground ground point cloud data;

[0164] removing a point data connected region in which the number of point data is lower than a number threshold from the determined point data connected region to obtain a first foreground point data connected region.

[0165] Optionally, the foreground point cloud data determining module 300 is configured to collect point data around a position of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data, and can specifically include:

[0166] collecting point data within a predetermined local distance around the position of the point cloud skeleton in the ground point cloud data to obtain the local ground point cloud data.

[0167] Optionally, the ground marking feature recognition module 400 is configured to recognize geometric features of a ground marking based on the foreground point cloud data, and can specifically include:

[0168] extracting point cloud data corresponding to the ground marking from the foreground point cloud data, and recognizing geometric features of the ground marking based on the point cloud data corresponding to the ground marking.

[0169] Optionally, the ground marker feature recognition module 400 is configured to extract the ground marker corresponding point cloud data from the foreground point cloud data, and can specifically include the following steps:

[0170] determining the distance between the point data and its nearest neighbor point data in the foreground point cloud data;

[0171] determining the connectivity distance threshold corresponding to the foreground point cloud data according to the average value and the standard deviation of the distance;

[0172] determining the point data connectivity region in which the point data in the foreground point cloud data is connected according to the connectivity distance threshold;

[0173] removing the point data connectivity region in which the number of point data is less than the number threshold, and collecting the remaining point data as the ground marker corresponding point cloud data.

[0174] The ground marker recognition device provided by the embodiment of the application can identify the ground marker corresponding point cloud data from the foreground part of the local ground according to the position and reflectivity information of the point cloud data, so as to realize the identification of the geometric features of the ground marker, avoid the information loss problem of the point cloud data caused by converting the point cloud data into a grid image, and accurately identify the geometric features of the ground marker.

[0175] The embodiment of the application also provides a computing device, which can execute the ground marker recognition method provided by the embodiment of the application by loading the ground marker recognition device in the form of a program. Figure 9 As shown in the figure, the computing device can include at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4.

[0176] In the embodiment of the application, the number of the processor 1, the communication interface 2, the memory 3 and the communication bus 4 is at least one, and the processor 1, the communication interface 2 and the memory 3 can communicate with each other through the communication bus 4.

[0177] Optionally, the communication interface 2 can be the interface of the communication module.

[0178] The processor 1 can be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiment of the application.

[0179] The memory 3 can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0180] The memory 3 stores a program, and the processor 1 invokes the program stored in the memory 3 to execute the ground mark recognition method provided by the embodiment of the application.

[0181] The embodiment of the application further provides a readable storage medium, which can store the program for executing the ground mark recognition method provided by the embodiment of the application.

[0182] Optionally, the program can be used for:

[0183] determining ground point cloud data;

[0184] determining a point cloud skeleton from the ground point cloud data; the reflectivity of point data in the point cloud skeleton is higher than a reflectivity threshold;

[0185] taking point cloud data around a position of the point cloud skeleton in the ground point cloud data as local ground point cloud data, and determining foreground point cloud data from the local ground point cloud data;

[0186] recognizing geometric features of the ground mark based on the foreground point cloud data.

[0187] Optional implementation and extended implementation of the program can refer to the description of the corresponding part in the foregoing description, and will not be described here.

[0188] Although the embodiments of the application are disclosed as above, the application is not limited to this. Any person skilled in the art, without departing from the spirit and scope of the application, can make various modifications and changes, therefore the protection scope of the application should be subject to the scope defined by the claims.

Claims

1. A method for recognizing ground markings, wherein, include: Determine the ground point cloud data; Determine the point cloud skeleton from ground point cloud data; The reflectance of the point data in the point cloud skeleton is higher than the reflectance threshold. Collect the point data within a predetermined local distance around the location of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data; Filter the point data in the local ground point cloud data whose reflectivity is lower than the foreground reflectivity threshold to determine the first local foreground ground point cloud data corresponding to the foreground part. From the first local foreground ground point cloud data, determine the first foreground point data connected region where the point data are connected, wherein the number of point data in the first foreground point data connected region is not less than the number threshold. The foreground cloud data is determined based on the first foreground data connectivity region; Based on the foreground point cloud data, the geometric features of the ground markings are identified.

2. The ground marking identification method according to claim 1, wherein, The determined ground point cloud data includes: Acquire the collected point cloud data; The corresponding ground point cloud data is determined from the collected point cloud data.

3. The ground marker identification method according to claim 2, wherein determining the corresponding ground point cloud data of the ground portion from the collected point cloud data includes: Determine the connected regions of point data from the collected point cloud data; The region with the most points among the identified connected regions is defined as the ground point cloud data.

4. The method for identifying ground markings according to claim 3, wherein, The process of determining the connected regions of point data from the collected point cloud data includes: Determine the distance between each point in the collected point cloud data, as well as the rate of elevation change between adjacent points; A set of point data whose elevation change rate between adjacent point data is less than the elevation change rate threshold and whose distance between point data is not greater than a set distance threshold is considered as a connected region of the same point data.

5. The method for identifying ground markings according to claim 1, wherein, Determining the point cloud skeleton from the ground point cloud data includes: The reflectance of the point data in the ground point cloud data is subjected to median filtering. The reflectance of the point data after median filtering is subjected to Laplace filtering. Filter out point data with reflectivity lower than the set reflectivity threshold from the point data after Laplace filtering; The point cloud skeleton is obtained by identifying connected regions of point data in the filtered point data and removing connected regions of point data whose number of point data is lower than a certain threshold.

6. The method for identifying ground markings according to claim 5, wherein, Median filtering of the reflectance of point data in the ground point cloud data includes: Determine the intermediate reflectance of the neighboring point data corresponding to the point data in the ground point cloud data, and use the intermediate reflectance as the reflectance of the point data.

7. The method for identifying ground markings according to claim 6, wherein, Determining the intermediate reflectance of neighboring point data corresponding to point data in the ground point cloud data includes: For any point in the ground point cloud data, determine the neighboring points of that point from the ground point cloud data; Sort the reflectance of the identified neighboring data points; If the number of determined neighboring data points is odd, the reflectance of the middle position corresponding to the sorting is determined as the middle reflectance; If the number of neighboring data points is even, the average of the two reflectivities in the middle is determined as the intermediate reflectivity.

8. The method for identifying ground markings according to claim 5, wherein, The Laplacian filtering process for the reflectance of the point data after median filtering includes: For any data point after median filtering, determine the neighboring data points of that data point. Determine the reflectance difference between the data point and each of its neighboring data points. Add up the determined reflectance differences and divide by the number of neighboring data points of the data point. Use the absolute value of the calculation result as the reflectance of the data point.

9. The method for identifying ground markings according to claim 1, wherein, If there are two or more point cloud skeletons, the method further includes: Merge point cloud skeletons whose distance between them is less than the predetermined merging distance.

10. The method according to claim 1, wherein, The step of determining the foreground cloud data based on the first foreground data connectivity region includes: Collect the point data within a predetermined local distance around the location of the first foreground point data connected region in the ground point cloud data; From the point data of the set, determine the second local foreground ground point cloud data corresponding to the foreground part, and from the second local foreground ground point cloud data, determine the second foreground point data connected region where the point data are connected, wherein the number of point data in the second foreground point data connected region is not less than the number threshold. If the difference in the number of point data between the second foreground point data connected region and the first foreground point data connected region is within a predetermined error, the first foreground point data connected region is determined as foreground point cloud data. If the difference in the number of point data between the second foreground point data connected region and the first foreground point data connected region is not within a predetermined error, the first foreground point data connected region is removed.

11. The method for identifying ground markings according to claim 1, wherein, The step of determining the first local foreground ground point cloud data corresponding to the foreground portion from the local ground point cloud data includes: Determine the foreground reflectance threshold for local ground point cloud data; Filter out point data in the local ground point cloud data whose reflectance is lower than the foreground reflectance threshold to obtain the first local foreground ground point cloud data; The first foreground point data connectivity region determined from the first local foreground ground point cloud data includes: Determine the connected regions of point data in the first local foreground ground point cloud data; Remove the connected regions of point data whose number of points is below the threshold to obtain the first connected regions of point data.

12. The method for identifying ground markings according to claim 1, wherein, The step of collecting point data within a predetermined local distance around the location of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data includes: From the ground point cloud data, the point data within a predetermined local distance around the location of the point cloud skeleton are collected to obtain the local ground point cloud data.

13. The method for identifying ground markings according to claim 1, wherein, The identification of geometric features of ground markers based on the foreground point cloud data includes: Extract the corresponding point cloud data of the ground markers from the foreground point cloud data, and identify the geometric features of the ground markers based on the corresponding point cloud data of the ground markers.

14. The method for identifying ground markings according to claim 13, wherein, The step of extracting the point cloud data corresponding to the ground markers from the foreground point cloud data includes: Determine the distance between the midpoint data of the foreground point cloud data and its nearest neighbor data; Based on the average value and standard deviation of the distance, determine the connectivity distance threshold corresponding to the foreground point cloud data; Based on the connectivity distance threshold, determine the point data connectivity region in the foreground point cloud data; Remove the connected regions of the identified point data where the number of point data is below the threshold, and collect the remaining point data into the corresponding point cloud data for ground marking.

15. A ground marking identification device, wherein, include: The ground point cloud determination module is used to determine ground point cloud data; A point cloud skeleton determination module is used to determine the point cloud skeleton from the ground point cloud data; The reflectance of the point data in the point cloud skeleton is higher than the reflectance threshold. The foreground point cloud data determination module is used to collect point data within a predetermined local distance around the location of the point cloud skeleton in the ground point cloud data to obtain local ground point cloud data; filter point data in the local ground point cloud data whose reflectivity is lower than the foreground reflectivity threshold to determine the first local foreground ground point cloud data corresponding to the foreground portion; determine the first foreground point cloud data connectivity region from the first local foreground point cloud data, wherein the number of point data in the first foreground point cloud data connectivity region is not less than a number threshold; and determine the foreground point cloud data based on the first foreground point cloud data connectivity region. The ground marker feature recognition module is used to identify the geometric features of ground markers based on the foreground point cloud data.

16. A readable storage medium, wherein, The readable storage medium stores a program for performing the ground marking identification method according to any one of claims 1-14.

Images