Modeling method, device and equipment of rod-shaped map element and storage medium

Abstract

Embodiments of the present disclosure disclose a rod-shaped map element modeling method and device, equipment and a storage medium, the method comprises: acquiring rod-shaped point cloud data; along the longitudinal direction of the rod-shaped point cloud data, the rod-shaped point cloud data is segmented, and at least two first point cloud data are obtained; based on each first point cloud data, the parameters of the first fitting circle corresponding to each first point cloud data are fitted; the center of the first fitting circle is linearly fitted to obtain the rod center axis of the rod-shaped map element; based on the rod center axis and the rod-shaped point cloud data, the rod-shaped map element corresponding to the rod-shaped object is modeled. The method can obtain the rod center axis close to the real rod-shaped object by segmenting and fitting the rod-shaped point cloud data, and then model the entire rod-shaped map element model based on the rod center axis and the rod-shaped point cloud data, so as to obtain the outer surface data close to the real rod-shaped object, thereby obtaining model data with high precision and good robustness.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of high-definition map, in particular to a modeling method and device of a rod-shaped map element, equipment and a storage medium. BACKGROUND

[0002] The rod-shaped map element is an important component of the high-definition map, which can correspond to a rod-shaped object element such as a traffic sign pole in the real world. The rod-shaped object element such as the traffic sign pole is also an important reference for positioning in the intelligent driving scene. Therefore, modeling the rod-shaped map element is an indispensable technical link for high-definition map and intelligent driving scene.

[0003] However, when the point cloud collection vehicle collects road point cloud information, the following situations exist for scanning the rod-shaped object: due to the influence of the shape of the rod-shaped object, the point cloud collection vehicle can only scan one side of the rod-shaped object, and cannot perform complete scanning; the bottom of the rod-shaped object is easily blocked by other vehicles or traffic barriers and other obstacles, resulting in incomplete scanning; the shape of the rod-shaped object may not be a true cylinder, and often the top is thin and the bottom is wide, and other objects are occasionally mounted in the middle, resulting in uneven point cloud density and different coverage degrees from the bottom to the top of the rod-shaped object scanned by the point cloud collection vehicle.

[0004] The above situations increase the difficulty of modeling the rod-shaped map element model of the rod-shaped object, thereby causing poor fitting accuracy and robustness of the rod-shaped map element model obtained by the traditional vector fitting method.

[0005] Therefore, it is necessary to propose a solution to solve the technical problem of poor fitting accuracy and robustness of the rod-shaped map element model in the high-definition map. SUMMARY

[0006] The present disclosure provides a modeling method, device, equipment and storage medium of a rod-shaped map element.

[0007] In a first aspect, a modeling method of a rod-shaped map element is provided in the embodiments of the present disclosure, and the method comprises:

[0008] obtaining rod-shaped object point cloud data;

[0009] segmenting the rod-shaped object point cloud data along a longitudinal direction of the rod-shaped object point cloud data to obtain at least two pieces of first point cloud data;

[0010] fitting a first fitting circle corresponding to each piece of first point cloud data based on each piece of first point cloud data, wherein the parameters of the first fitting circle at least include a center;

[0011] performing linear fitting on the center of the first fitting circle to obtain a rod center axis of the rod-shaped map element;

[0012] modeling a rod-shaped map element corresponding to the rod based on the rod central axis and the rod point cloud data.

[0013] Further, the method further comprises:

[0014] projecting each segment of the first point cloud data onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the segment of the first point cloud data;

[0015] performing circular fitting on the two-dimensional point cloud data to obtain parameters of a first fitted circle corresponding to the first point cloud data, the parameters at least comprising a circle center.

[0016] Further, the projecting each segment of the first point cloud data onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the segment of the first point cloud data comprises:

[0017] setting a height of each segment of the first point cloud data as a height of the plane to obtain the corresponding two-dimensional point cloud data.

[0018] Further, the parameters further comprise an edge line of the first fitted circle.

[0019] After the fitting the parameters of the first fitted circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, the method further comprises:

[0020] determining a vector distance of each point cloud data in each segment of the first point cloud data to the edge line of the corresponding first fitted circle in a horizontal direction perpendicular to the longitudinal direction;

[0021] determining a fitting residual of the corresponding first fitted circle based on the vector distance corresponding to each segment of the first point cloud data;

[0022] eliminating the first fitted circle whose fitting residual does not satisfy a residual threshold condition from the plurality of first fitted circles.

[0023] Further, after the fitting the parameters of the first fitted circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, the method further comprises:

[0024] determining a point cloud coverage degree based on the two-dimensional point cloud data corresponding to the first point cloud data, the point cloud coverage degree being a ratio of a length covered by the edge line of the first fitted circle to a circumference of the first fitted circle.

[0025] remove the first fitting circle whose point cloud coverage does not satisfy the coverage threshold condition from the plurality of first fitting circles.

[0026] Further, the modeling of the corresponding rod-shaped map element of the rod-shaped object based on the rod center axis and the rod-shaped object point cloud data comprises:

[0027] fitting a second fitting circle corresponding to each segment of first point cloud data based on each segment of first point cloud data, wherein the center of the second fitting circle is located on the rod center axis;

[0028] determining the radius and the center of the second fitting circle corresponding to the first point cloud data as the modeling data of the rod-shaped map element.

[0029] Further, the modeling of the corresponding rod-shaped map element of the rod-shaped object based on the rod center axis and the rod-shaped object point cloud data comprises:

[0030] re-segmenting the rod-shaped object point cloud data in the longitudinal direction of the rod-shaped object point cloud data to obtain at least two segments of second point cloud data;

[0031] fitting a third fitting circle corresponding to each segment of second point cloud data based on each segment of second point cloud data, wherein the center of the third fitting circle is located on the rod center axis;

[0032] determining the radius and the center of the third fitting circle corresponding to the second point cloud data as the modeling data of the rod-shaped map element.

[0033] re-segmenting the rod-shaped object point cloud data to obtain at least two segments of second point cloud data.

[0034] In a second aspect, an embodiment of the present application provides a map making method, which models a rod-shaped map element in a map by using the method of the first aspect, and makes a map based on the modeled rod-shaped map element.

[0035] In a third aspect, an embodiment of the present application provides a modeling device of a rod-shaped map element, wherein the modeling device comprises:

[0036] The acquisition module is configured to acquire rod-shaped object point cloud data.

[0037] The segmentation module is configured to segment the rod-shaped object point cloud data in the longitudinal direction of the rod-shaped object point cloud data to obtain at least two segments of first point cloud data.

[0038] The first fitting module is configured to fit a first fitting circle corresponding to each segment of first point cloud data based on each segment of first point cloud data, wherein the parameters of the first fitting circle at least include a center.

[0039] a second fitting module, configured to perform a straight line fitting on the center of the first fitting circle to obtain a rod center axis of the rod-shaped map element;

[0040] a modeling module, configured to model a rod-shaped map element corresponding to the rod-shaped object based on the rod center axis and the rod-shaped object point cloud data.

[0041] The functions can be implemented by hardware, or by hardware executing software in the relevant field. The hardware or software includes one or more modules corresponding to the functions described above.

[0042] In one possible design, the apparatus includes a memory and a processor. The memory is configured to store one or more computer instructions for supporting the apparatus to perform the corresponding method described above. The processor is configured to execute the computer instructions stored in the memory. The apparatus can further include a communication interface configured to enable the apparatus to communicate with other devices or communication networks.

[0043] In a fourth aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the method of any of the aspects described above.

[0044] In a fifth aspect, a computer-readable storage medium is provided, which is configured to store computer instructions for the apparatus described above. The computer instructions are executed by a processor to implement the method of any of the aspects described above.

[0045] In a sixth aspect, a computer program product is provided, which includes computer instructions. The computer instructions are executed by a processor to implement the method of any of the aspects described above.

[0046] The technical solutions provided by the embodiments of the present disclosure can have the following beneficial effects:

[0047] The modeling method of the rod-shaped map element provided by the embodiments of the present disclosure first divides the collected rod-shaped point cloud data into multiple sections of point cloud data in the longitudinal direction. Then, each section of the multiple sections of point cloud data is respectively subjected to circle fitting, that is, each section of point cloud data is fitted into a circle. After that, the centers of the multiple fitted circles are subjected to straight line fitting to obtain the rod center axis of the rod-shaped map element, and then the rod-shaped map element is modeled based on the rod center axis and the rod-shaped point cloud data. The above modeling method provided by the embodiments of the present disclosure, by segmenting and fitting the rod-shaped point cloud data, since the centers of most of the circles obtained by the segmented fitting are close to the rod center axis of the real rod-shaped object, the straight line fitting based on the centers of the circles can obtain a rod center axis with higher accuracy. After that, the rod-shaped map element is subjected to segmented modeling, each section of the modeling data corresponds to a circle whose center is located on the rod center axis, and the radius of the circle is determined by the rod-shaped point cloud data corresponding to the section of modeling data. Therefore, the segmented modeling data can obtain the outer surface data close to the real rod-shaped object, so as to obtain the model data of the rod-shaped map element with high precision and good robustness.

[0048] It should be understood that the general description above and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0049] Other features, objects, and advantages of the present disclosure will become more apparent from the following detailed description of the non-limiting embodiments, taken in conjunction with the accompanying drawings. In the drawings:

[0050] Figure 1 A flow chart of a modeling method of a rod-shaped map element according to an embodiment of the present disclosure is shown;

[0051] Figure 2 An implementation schematic diagram of modeling of a rod-shaped map element according to an embodiment of the present disclosure is shown;

[0052] Figure 3 A structural block diagram of a modeling device of a rod-shaped map element according to an embodiment of the present disclosure is shown;

[0053] Figure 4 A structural schematic diagram of an electronic device suitable for implementing the modeling method of a rod-shaped map element according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0054] Hereinafter, the exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily implement them. In addition, parts irrelevant to the description of the exemplary embodiments are omitted in the drawings for the sake of clarity.

[0055] In this disclosure, it should be understood that terms such as “comprising” or “having” are intended to indicate the presence of features, figures, steps, behaviors, components, parts or combinations thereof disclosed in this specification, and do not preclude the possibility of the presence or addition of one or more other features, figures, steps, behaviors, components, parts or combinations thereof.

[0056] It should also be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other. This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0057] The details of the embodiments of this disclosure are described in detail below through specific examples.

[0058] Figure 1 A flowchart illustrating a method for modeling bar-shaped map elements according to an embodiment of this disclosure is shown. Figure 1 As shown, the modeling method for this bar-shaped map element includes the following steps:

[0059] In step S101, point cloud data of the rod-shaped object is acquired;

[0060] In step S102, the rod-shaped object point cloud data is segmented along the longitudinal direction to obtain at least two segments of first point cloud data;

[0061] In step S103, based on each segment of the first point cloud data, the parameters of the first fitted circle corresponding to each segment of the first point cloud data are fitted.

[0062] The parameters include at least the center of the circle;

[0063] In step S104, a straight line is fitted to the center of the first fitted circle to obtain the central axis of the rod-shaped map element;

[0064] In step S105, based on the central axis of the pole and the point cloud data of the pole-shaped object, the pole-shaped map element corresponding to the pole-shaped map element is modeled.

[0065] In this embodiment, the modeling method for the pole-shaped map element can be executed on a map production server. The pole-shaped map element can be a model element in a high-precision map that corresponds to a pole-shaped object in the real world. For example, the pole-shaped map element can correspond to traffic sign poles in the real world, such as street lamp poles, camera poles, traffic light poles, and sign poles.

[0066] Before modeling the pole-shaped map element, a pole-shaped point cloud data of a pole-shaped object such as a traffic sign pole appearing on both sides of a real road can be collected by a point cloud collection vehicle. It should be noted that the pole-shaped point cloud data in the embodiments of the present disclosure can include but is not limited to point cloud data collected by a laser radar loaded on the point cloud collection vehicle, and visual point cloud data generated after image processing of an image collected by an image collection device installed on the point cloud collection vehicle.

[0067] In the related art, based on the pole-shaped point cloud data, a least square, a fitting algorithm (Random Sample Consensus, RANSAC), a Hough transform, or the like is used to perform fitting of a pole-shaped map element model to obtain modeling data of the pole-shaped map element. However, since the pole-shaped object in reality is not a strict cylindrical model, the shape is often thin at the top and wide at the bottom, and occasionally has other objects attached in the middle, that is, an irregular shape of a columnar object with a thin top and a thick bottom, and other uncertain shape attachments, so that the overall fitting method of the pole-shaped map element model is affected by the irregular shape of the pole-shaped map element and the noise of the attachments thereon, thereby resulting in low fitting accuracy and poor robustness of the existing fitting method. Therefore, the embodiments of the present disclosure propose a modeling method of a pole-shaped map element.

[0068] In some embodiments, the modeling method of the pole-shaped map element provided by the embodiments of the present disclosure can be used in the scenario of modeling the pole-shaped map element in a high-definition map.

[0069] In some embodiments, the pole-shaped point cloud data can be obtained by scanning the surface of the pole-shaped object on both sides of the road by a laser or other data collection device when the point cloud collection vehicle collects road information.

[0070] In some embodiments, the pole-shaped point cloud data can be a data form used to represent the surface information of the pole-shaped object in a three-dimensional space, and can be specifically represented as the three-dimensional coordinates (such as longitude, latitude, and height) of a certain position point on the surface of the pole-shaped object.

[0071] In some embodiments, the pole-shaped point cloud data can be divided into at least two segments of first point cloud data along the longitudinal direction of the pole-shaped point cloud data, and each segment is a group of first point cloud data. In some embodiments, the pole-shaped point cloud data can be divided according to the same or different heights. That is, the pole-shaped point cloud data can be evenly divided into multiple segments of first point cloud data with the same height, or the pole-shaped point cloud data can be divided into multiple segments of first point cloud data with different heights along the longitudinal direction of the pole-shaped point cloud data. The specific division can be determined according to actual conditions, and the embodiments of the present disclosure do not limit the same. It should be understood that the longitudinal direction in the embodiments is a direction perpendicular to the horizontal direction, and can also be understood as the height direction of the pole-shaped object.

[0072] In some embodiments, each of the plurality of segments of the first point cloud data can be fitted into a corresponding first fitted circle respectively, and the plurality of segments of the first point cloud data can correspond to parameters of the plurality of first fitted circles, the parameters at least including the center of the circle. In some embodiments, since each point in each segment of the first point cloud data has the same or different height, a two-dimensional point can be obtained by projecting all three-dimensional points in the segment of the first point cloud data onto a plane perpendicular to the longitudinal direction, and then a corresponding first fitted circle and the center of the first fitted circle can be fitted based on the two-dimensional point. The fitting algorithm may, for example, adopt the RANSAC algorithm, and it can be understood that other fitting algorithms can also be adopted, and the embodiments of the present disclosure are not limited thereto.

[0073] In some embodiments, after fitting at least two segments of the first point cloud data obtained by segmenting the rod-shaped point cloud data into first fitted circles respectively, a fitting algorithm such as the RANSAC algorithm can be used to perform linear fitting on the centers of the plurality of first fitted circles, and the fitted straight line can be used as the rod center axis of the rod. By this way of first segmenting and fitting to obtain a circle, and then fitting a straight line based on the centers of the circles to obtain the rod center axis of the rod, noise caused by irregular shapes such as attachments on the rod and rod branches on the fitting of the rod-shaped map element model can be filtered, and good robustness can be achieved.

[0074] In some embodiments, after the rod center axis of the rod-shaped object is determined, a rod-shaped map element corresponding to the rod-shaped object can be modeled based on the rod center axis and the rod-shaped point cloud data. In some embodiments, the rod-shaped map element can be modeled as a columnar model with the rod center axis as the center line, and the outer surface of the columnar model is modeled based on the collected rod-shaped point cloud data, so that a real shape close to the corresponding rod-shaped object in the real world can be fitted.

[0075] The modeling method of the rod-shaped map element provided in the embodiments of the present disclosure first divides the collected rod-shaped point cloud data into multiple segments of point cloud data in the longitudinal direction; then performs circular fitting on each segment of the multiple segments of point cloud data, that is, fitting each segment of point cloud data into a circle; then performs straight line fitting on the centers of the multiple fitted circles to obtain the rod center axis of the rod-shaped object, and then models the rod-shaped map element based on the rod center axis and the rod-shaped point cloud data. The modeling method provided in the embodiments of the present disclosure performs segmented circular fitting on the rod-shaped point cloud data, and since the centers of most of the circles obtained by the segmented fitting are close to the rod center axis of the real rod-shaped object, the straight line fitting based on the centers can obtain a rod center axis with higher accuracy, and then the segmented modeling is performed on the rod-shaped map element, each segment of modeling data corresponds to a circle whose center is located on the rod center axis, and the radius of the circle is determined by the rod-shaped point cloud data corresponding to the segment of modeling data, so that the segmented modeling data can obtain the outer surface data close to the real rod-shaped object, thereby obtaining the model data of the rod-shaped map element with high precision and good robustness.

[0076] In an optional implementation of the embodiments, the step S103, that is, fitting the parameters of the first fitted circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, the parameters at least including the center of the circle, can further include the following steps:

[0077] projecting each segment of the first point cloud data onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the segment of the first point cloud data;

[0078] performing circular fitting on the two-dimensional point cloud data to obtain the parameters of the first fitted circle corresponding to the first point cloud data, the parameters at least including the center of the circle.

[0079] In the optional implementation, as described above, the first point cloud data is a three-dimensional point cloud, and after projecting all or part of the first point cloud data on each segment obtained by the division onto a plane perpendicular to the longitudinal direction, a set of two-dimensional point clouds located on the plane can be obtained, that is, the three-dimensional point cloud on each segment obtained by the division is projected onto a plane, and then a circular fitting algorithm is used to perform circular fitting on the obtained two-dimensional point cloud data to obtain a first fitted circle, each segment of the first point cloud data corresponding to a first fitted circle. After obtaining multiple first fitted circles, the parameters of each first fitted circle can also be obtained, for example, the parameters including the center of the circle. In some embodiments, the circular fitting algorithm is, for example, the RANSAC algorithm.

[0080] In some embodiments, the first point cloud data of different segments can be projected onto the same projection plane or different projection planes.

[0081] In an optional implementation of the embodiment, the step of projecting each segment of the first point cloud data onto a plane perpendicular to the longitudinal direction to obtain the corresponding two-dimensional point cloud data of the segment of the first point cloud data can further include the following steps:

[0082] Setting the height of each segment of the first point cloud data as the height of the plane to obtain the corresponding two-dimensional point cloud data.

[0083] In some embodiments, the height of the same segment of the first point cloud data can be set as the height of a plane, such as 0 or other numerical values.

[0084] It can be understood that, in the process of segment fitting, for each segment of the first point cloud data, the density of the point cloud can be greatly increased by compressing the three-dimensional point cloud to a two-dimensional space, thereby overcoming the problem of inaccurate fitting caused by sparse point cloud.

[0085] The embodiments of the present disclosure, when segment fitting, compress the first point cloud data of each segment from a three-dimensional point cloud to a two-dimensional space before fitting the circle, rather than directly fitting a cylinder using each segment of the first point cloud data obtained by segmentation, thereby reducing the number of fitting parameters and lowering the difficulty of fitting. For the case of fewer points, it has better robustness.

[0086] In an optional implementation of the embodiment, the parameters further include an edge line of the first fitting circle; after step S103, i.e., after the step of fitting the parameters of the first fitting circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, the method can further include the following steps:

[0087] Determining the vector distance of each point cloud data in each segment of the first point cloud data to the edge line of the corresponding first fitting circle in the horizontal direction perpendicular to the longitudinal direction;

[0088] Determining the fitting residual of the corresponding first fitting circle based on the vector distance corresponding to each segment of the first point cloud data;

[0089] Removing the first fitting circle that does not satisfy the residual threshold condition from the plurality of first fitting circles.

[0090] In the optional implementation, as described above, the rod-like object on the real road is generally not a regular cylinder, and some attachments can be present thereon, and the center of the first fitting circle obtained by fitting the segmented circles is not necessarily on the actual rod center axis of the rod-like object. If the actual rod center axis of the rod-like object is determined based on the centers of the first fitting circles, more noise can be brought, thereby causing the fitting rod center axis to deviate greatly from the actual rod center axis of the rod-like object. Therefore, the embodiment needs to filter out noise from the obtained first fitting circles after the first fitting circles are fitted, that is, the first fitting circles with the center of which is not on the actual rod center axis of the rod-like object are removed, and the first fitting circles with the center of which is on the actual rod center axis of the rod-like object are retained and subjected to the next straight line fitting.

[0091] In some embodiments, the noise filtering is performed by the fitting residual of the first fitting circle. The fitting residual threshold can be determined in advance. When the fitting residual of the first fitting circle is greater than the predetermined fitting residual threshold, it is considered that the first fitting circle does not meet the above requirement and is filtered out as noise.

[0092] In some embodiments, the fitting residual of the first fitting circle can be calculated as follows: in the circle fitting, for the first point cloud data of the same segment, each point cloud data in the first point cloud data of the same segment is projected on a plane perpendicular to the longitudinal direction to obtain a set of two-dimensional point cloud data, and the circle fitting is performed based on the set of two-dimensional point cloud data to obtain the first fitting circle. In the embodiment, after the first fitting circle is fitted, the vector distance of each two-dimensional point cloud data in the set of two-dimensional point cloud data to the edge line of the first fitting circle can be determined, and the fitting residual of the first fitting circle can be obtained by adding the vector distances corresponding to the set of two-dimensional point cloud data and dividing the sum by the number of point clouds in the set of two-dimensional point cloud data. It can be understood that the fitting residual of the first fitting circle is not limited to the above calculation method, and the embodiments of the present disclosure are not limited in this regard.

[0093] In some embodiments, the fitting residual threshold can be set to 2 cm, 1 cm or other values, and the embodiments of the present disclosure are not limited in this regard.

[0094] In some embodiments, the fitting residual greater than or equal to the fitting residual threshold can be understood as the fitting residual not meeting the residual threshold condition.

[0095] In the optional implementation, after obtaining the fitting residual of each first fitting circle, it is determined whether the corresponding first fitting circle is qualified by judging whether the fitting residual satisfies a residual threshold condition; when the fitting residual does not satisfy the residual threshold condition, it is determined that the first fitting circle is unqualified, so that the first fitting circle can be removed from the plurality of first fitting circles, and the remaining first fitting circles are retained, that is, the remaining first fitting circles are qualified. In this way, the first fitting circles that meet the requirements can be screened out by removing the unqualified first fitting circles according to the fitting residual of the first fitting circles, so that the noise interference is reduced and the fitting accuracy is improved.

[0096] In an optional implementation of the embodiment, after step S103, that is, after the parameters of the first fitting circle corresponding to each segment of first point cloud data are fitted based on each segment of first point cloud data, the method can further include the following steps:

[0097] Based on the two-dimensional point cloud data corresponding to the first point cloud data, a point cloud coverage degree is determined according to a ratio of a length covered by an edge line of the first fitting circle to a circumference of the first fitting circle;

[0098] The first fitting circle whose point cloud coverage degree does not satisfy a coverage degree threshold condition is removed from the plurality of first fitting circles.

[0099] In the optional implementation, as described above, the rod-shaped object on the real road is usually not a regular cylinder, and there can be some attachments thereon. The center of the first fitting circle obtained by fitting the segmented circle is not necessarily on the actual rod center axis of the rod-shaped object. If the actual rod center axis of the rod-shaped object is determined by fitting a straight line based on the centers of the first fitting circles, more noise can be brought, so that the fitting obtained rod center axis deviates greatly from the actual rod center axis of the rod-shaped object. Therefore, after the first fitting circle is obtained, the embodiment needs to filter out noise from the obtained first fitting circle, that is, to remove the first fitting circle whose center is not on the actual rod center axis of the rod-shaped object, and to retain the first fitting circle whose center is on the actual rod center axis of the rod-shaped object, and then to perform the next step of straight line fitting.

[0100] In some embodiments, by filtering out noise from the point cloud coverage degree of the fitted first fitting circle through the point cloud data, a coverage degree threshold can be determined in advance. When the point cloud coverage degree of the first fitting circle is less than or equal to the predetermined coverage degree threshold, it is considered that the first fitting circle does not meet the above requirements and is filtered out as noise.

[0101] In some embodiments, the point cloud coverage degree of the first fitting circle can be understood as a ratio of a length covered by an edge line of the first fitting circle to a circumference of the first fitting circle, with respect to the two-dimensional point cloud data used to fit the first fitting circle. As shown in FIG. 4, the point cloud coverage degree of the first fitting circle is the ratio of the length covered by the edge line of the first fitting circle to the circumference of the first fitting circle.Figure 2 As shown, the coverage relationship between the point cloud data and the first fitted circle is shown in the middle lower area, assuming that the length of the point cloud data covered along the edge line of the first fitted circle is L, the circumference of the first fitted circle is 2pR based on the radius R of the first fitted circle, and the point cloud coverage of the first fitted circle is L / 2pR. In this embodiment, the coverage threshold can be set to 1 / 2, 2 / 3 or other possible values, which is determined according to the actual modeling accuracy requirement, and the embodiments of the present disclosure are not limited in this regard.

[0102] In other optional embodiments, the point cloud coverage of the first fitted circle can also be understood as the coverage rate of the two-dimensional point cloud data used to fit the first fitted circle on the circumference of the first fitted circle. For example, the point cloud coverage of the first fitted circle can be determined as the number of degrees of the circumference of the first fitted circle covered by the two-dimensional point cloud data. As shown, the coverage relationship between the point cloud data and the first fitted circle is shown in the middle lower area, and the point cloud data covers more than half of the circumference of the first fitted circle, that is, the point cloud coverage of the first fitted circle is more than 180 degrees. In some embodiments, the coverage threshold can be set to 120 degrees, 180 degrees or other possible degrees, which is determined according to the actual modeling accuracy requirement, and the embodiments of the present disclosure are not limited in this regard. Figure 2

[0103] In some embodiments, the point cloud coverage less than or equal to the coverage threshold can be understood as that the point cloud coverage does not meet the coverage threshold condition.

[0104] In this optional implementation, after obtaining the point cloud coverage of each first fitted circle, it is determined whether the corresponding first fitted circle is qualified by judging whether the point cloud coverage meets the coverage threshold. When the point cloud coverage does not meet the coverage threshold, it can be determined that the first fitted circle is unqualified, so that the first fitted circle is removed from the plurality of first fitted circles, and the remaining first fitted circles are retained, that is, the remaining first fitted circles are qualified. This way of removing unqualified first fitted circles through the fitting residual of the first fitted circle can filter out the qualified first fitted circles, thereby reducing noise interference and improving fitting accuracy.

[0105] In some embodiments, the first fitted circle can be removed from the plurality of first fitted circles when the fitting residual of the first fitted circle does not meet the residual threshold condition and / or the point cloud coverage of the first fitted circle does not meet the coverage threshold condition, so that the center axis of the rod-shaped object is obtained by using the center of the remaining first fitted circle for straight line fitting.

[0106] ​In an optional implementation of the embodiment, the step S105 of modeling the rod-shaped map element corresponding to the rod-shaped object based on the rod center axis and the rod-shaped object point cloud data can further include the following steps:

[0107] fitting a second fitting circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, the center of the second fitting circle being located on the rod center axis;

[0108] determining the radius and the center of the second fitting circle corresponding to the first point cloud data as the modeling data of the rod-shaped map element.

[0109] In the optional implementation, after the rod center axis of the rod-shaped map element is fitted, the modeling data of the rod-shaped map element can be fitted by using the collected first point cloud data and the rod center axis. To this end, in some embodiments, circular fitting can be performed again for each segment of the at least two segments of the first point cloud data obtained by segmentation. When the circle is fitted again, the center of the circle can be set on the rod center axis in advance, and a corresponding second fitting circle can be fitted based on a circle fitting algorithm such as the least square method. All the second fitting circles obtained in this way have a point on the rod center axis as the center, and the radius of the circle depends on the distribution of the segment of the first point cloud data, which is roughly consistent with the cylindrical surface of the rod-shaped object on the real road. It should be noted that when the second fitting circle is fitted, the intersection of the rod center axis and the projection plane after the segment of the first point cloud data is projected onto the same projection plane can be taken as the center of the second fitting circle corresponding to the segment of the first point cloud data, and then the radius of the second fitting circle can be fitted based on the center and the segment of the first point cloud data. Finally, the second fitting circles corresponding to all the first point cloud data and the radius of each second fitting circle can be taken as the modeling data of the rod-shaped map element, so that an element model consistent with the shape of the corresponding rod-shaped object on the real road can be rendered when the map is rendered.

[0110] In an optional implementation of the embodiment, the step S105 of modeling the rod-shaped map element corresponding to the rod-shaped object based on the rod center axis and the rod-shaped object point cloud data can further include the following steps:

[0111] re-segmenting the rod-shaped object point cloud data in the longitudinal direction of the rod-shaped object point cloud data to obtain at least two segments of second point cloud data;

[0112] fitting a third fitting circle corresponding to each segment of the second point cloud data based on each segment of the second point cloud data, the center of the third fitting circle being located on the rod center axis;

[0113] The radius and the center of the third fitting circle corresponding to the second point cloud data are determined as the modeling data of the rod-shaped map element.

[0114] In this optional implementation, after the rod center axis of the rod-shaped map element is fitted, the modeling data of the rod-shaped map element can be fitted by using the first point cloud data collected and the rod center axis. To this end, in some embodiments, the rod-shaped point cloud data can be re-divided in the longitudinal direction of the rod-shaped point cloud data, and the heights of the multiple pieces of second point cloud data obtained after the division can be the same or different. Then, for each piece of second point cloud data obtained after the re-division, a circular fitting is performed, and when the circle is fitted, the center of the circle can be preset on the rod center axis, and a corresponding third fitting circle is fitted based on a circle fitting algorithm such as the least square method. All the third fitting circles obtained in this way have a point on the rod center axis as the center, and the radius of the circle depends on the distribution of the piece of second point cloud data, which is roughly consistent with the cylindrical surface of the rod-shaped object on the real road. It should be noted that when the third fitting circle is fitted, the intersection of the rod center axis and the projection plane after the projection of a piece of second point cloud data to the same projection plane can be taken as the center of the third fitting circle corresponding to the piece of second point cloud data, and then the radius of the third fitting circle is fitted based on the center and the piece of second point cloud data. Finally, the third fitting circles corresponding to all the second point cloud data and the radius of each third fitting circle can be taken as the modeling data of the rod-shaped map element, so that an element model consistent with the shape of the corresponding rod-shaped object on the real road can be rendered when the map is rendered.

[0115] In this embodiment, the rod-shaped point cloud data can be re-divided based on different requirements for model fitting accuracy. It can be understood that the number of segments and the height of the multiple pieces of second point cloud data obtained by re-dividing the rod-shaped point cloud data are different from those of the multiple pieces of first point cloud data obtained by the previous division. When the model fitting accuracy requirement is high, the second point cloud data with a lower height can be divided, that is, compared with the process of dividing the multiple pieces of first point cloud data, the re-division of the multiple pieces of second point cloud data has a finer division granularity, more third fitting circles are fitted, and the model of the rod-shaped map element modeled based on the third fitting circles and the rod center axis has a higher accuracy; when the model fitting accuracy requirement is low, the second point cloud data with a higher height can be divided, that is, compared with the process of dividing the multiple pieces of first point cloud data, the re-division of the multiple pieces of second point cloud data has a larger division granularity, fewer third fitting circles are fitted, and the model of the rod-shaped map element modeled based on the third fitting circles and the rod center axis has a lower accuracy, but the calculation efficiency can be improved.

[0116] It should be noted that after fitting the central axis of the rod-shaped map element, whether one directly uses each segment of the first point cloud data obtained from the segmentation to perform circular fitting, or resegments the rod-shaped point cloud data to obtain multiple segments of the second point cloud data and performs circular fitting on each segment of the second point cloud data, the final modeling data for the rod-shaped map element can be obtained. Therefore, this segmented circular fitting method allows the rod-shaped map element to be represented as a multi-segment cylindrical model sharing a central axis, thus overcoming the problems of high fitting difficulty and poor fitting accuracy caused by the varying thickness of the upper and lower parts of rod-shaped objects in the real world.

[0117] The modeling method for pole-shaped map elements provided in the above embodiments enables a model fitting accuracy within 2cm for real-world pole-shaped objects such as signposts, streetlights, and camera poles, with a success rate as high as 98%.

[0118] According to one embodiment of this disclosure, a map production method is also proposed, which uses the above-mentioned modeling method for rod-shaped map elements to model rod-shaped map elements in a map, and produces a map based on the modeled rod-shaped map elements.

[0119] In this embodiment, pole-shaped objects, as essential traffic facilities on real roads, require corresponding pole-shaped map elements to be created on the electronic map. In this embodiment, pole-shaped objects on real roads can be modeled using the pole-shaped map element modeling method described above, and the modeling data corresponding to the pole-shaped map element is associated and stored with the modeling data of other map elements in the electronic map. During map rendering, the pole-shaped map element can be rendered in the current area of ​​the map page based on its modeling data.

[0120] For details on modeling bar map elements, please refer to the description of the modeling method for bar map elements above, which will not be repeated here.

[0121] Figure 2 This diagram illustrates one implementation of modeling a bar-shaped map element according to an embodiment of the present disclosure. Figure 2 As shown, the leftmost sub-image is a realistic illustration of a pole-shaped object on a real road. In this embodiment, a data acquisition vehicle scans the pole-shaped object to collect point cloud data. The point cloud data is then segmented into at least two segments along its longitudinal direction, such as... Figure 2The third subgraph shows the images corresponding to three segments of the first point cloud data, and the other segments are replaced by ellipses. In the fitting stage of the first fitting circle, the embodiment projects each segment of the first point cloud data obtained by the segmentation onto a plane to obtain a set of two-dimensional plane point clouds, and performs circular fitting on the set of two-dimensional plane point clouds by using the RANSAC method to obtain a corresponding first fitting circle. In order to filter out noise, the fitting residual and the point cloud coverage of each first fitting circle are also checked, and the first fitting circle that does not meet the requirements of the fitting residual and the point cloud coverage is removed, thereby obtaining a plurality of qualified first fitting circles as shown in the fifth subgraph. Figure 2 For the remaining qualified first fitting circles, the fitting of the rod center axis is performed again. In this process, the centers of all the remaining first fitting circles can be used to perform linear fitting to obtain the rod center axis. As can be seen from the fifth subgraph, the centers of a small number of first fitting circles are not on the same straight line as the centers of most first fitting circles, but the rod center axis obtained by fitting still substantially coincides with the actual rod center axis of the rod-shaped object. After the rod center axis is determined, the rod-shaped map element can be modeled based on the rod center axis and the rod-shaped object point cloud data, that is, the rod-shaped map element as shown in the rightmost part of the sixth subgraph is modeled. Figure 2 The centers and radii of the second fitting circles (or third fitting circles) can be recorded as modeling data of the rod-shaped map element in the map data, so that the shape of the rod-shaped map element consistent with the rod-shaped object on the real road can be rendered based on the modeling data when the map is rendered.

[0122] The following is an embodiment of the device of the present disclosure, which can be used to execute the method embodiments of the present disclosure.

[0123] Figure 3 A structural block diagram of a modeling device of a rod-shaped map element according to an embodiment of the present disclosure is shown. The device can be realized as part or all of an electronic device by software, hardware, or a combination of both. As shown in the figure, the modeling device includes: Figure 3

[0124] The acquisition module 301 is configured to acquire rod-shaped object point cloud data.

[0125] The segmentation module 302 is configured to segment the rod-shaped object point cloud data along the longitudinal direction of the rod-shaped object point cloud data to obtain at least two segments of first point cloud data.

[0126] The first fitting module 303 is configured to fit the parameters of a first fitting circle corresponding to each segment of the first point cloud data based on each segment of the first point cloud data, wherein the parameters at least include the center.

[0127] ​The second fitting module 304 is configured to perform a straight line fitting on the center of the first fitting circle to obtain a rod center axis of the rod-shaped map element.

[0128] The modeling module 305 is configured to model the rod-shaped map element based on the rod center axis and the rod-shaped point cloud data.

[0129] In this embodiment, the modeling device of the rod-shaped map element can be executed on a map making server. The rod-shaped map element can be a model element corresponding to a rod-shaped object in the real world in a high-definition map. For example, the rod-shaped map element can correspond to a traffic sign pole in the real world, such as a street lamp pole, a camera pole, a traffic light pole, a hanging plate pole, etc.

[0130] Before modeling the rod-shaped map element, the rod-shaped point cloud data of the rod-shaped object such as a traffic sign pole can be collected by a point cloud collection vehicle on both sides of the real road. It should be noted that the rod-shaped point cloud data in the embodiment of the present disclosure can include but is not limited to the point cloud data collected by the laser radar loaded on the point cloud collection vehicle, and the image collected by the image collection device installed on the point cloud collection vehicle, and the visual point cloud data generated after image processing.

[0131] In the related art, based on the rod-shaped point cloud data, a least square, a fitting algorithm (Random Sample Consensus, RANSAC), a Hough transform, etc. are used to fit the model of the rod-shaped map element to obtain the modeling data of the rod-shaped map element. However, since the rod-shaped object in the real world is not a strict cylindrical model, its shape is often thin at the top and wide at the bottom, and occasionally has other objects attached in the middle, that is, it is an irregular shape of a column with a thin top and a thick bottom, and other uncertain attachments, which affects the overall fitting device of the rod-shaped map element model due to the irregular shape of the rod-shaped map element and the noise of the attachments thereon, thereby resulting in low fitting accuracy and poor robustness of the existing fitting device. Therefore, the present embodiment provides a modeling device of a rod-shaped map element.

[0132] In some embodiments, the modeling device of the rod-shaped map element provided by the present embodiment can be used in the scene of modeling the rod-shaped map element in the high-definition map.

[0133] In some embodiments, the rod-shaped point cloud data can be obtained by scanning the surface of the rod-shaped object on both sides of the road by a laser data collection device when the point cloud collection vehicle collects road information.

[0134] In some embodiments, the rod-shaped point cloud data can be a data form used to represent the surface information of the rod-shaped object in a three-dimensional space, which can be specifically represented as the three-dimensional coordinates (such as longitude, latitude, and height) of a certain position point on the surface of the rod-shaped object.

[0135] In some embodiments, the rod point cloud data can be segmented into at least two pieces of first point cloud data along a longitudinal direction of the rod point cloud data, each piece being a group of first point cloud data. In some embodiments, the rod point cloud data can be segmented at the same or different heights. That is, the rod point cloud data can be evenly segmented into multiple pieces of first point cloud data of the same height, or segmented into multiple pieces of first point cloud data of different heights along the longitudinal direction of the rod point cloud data, which can be determined according to actual conditions, and the embodiments of the present disclosure are not limited thereto. It should be understood that the longitudinal direction in this embodiment is a direction perpendicular to the horizontal direction, and can also be understood as a height direction.

[0136] In some embodiments, each piece of the multiple pieces of first point cloud data can be fitted into a corresponding first fitted circle, and the multiple pieces of first point cloud data can correspond to parameters of the multiple first fitted circles, the parameters at least including a center. In some embodiments, since each point in each piece of first point cloud data has the same or different height, a two-dimensional point can be obtained by projecting all three-dimensional points in the piece of first point cloud data onto a plane perpendicular to the longitudinal direction, and then a corresponding first fitted circle and a center of the first fitted circle can be fitted based on the two-dimensional point. The fitting algorithm may, for example, adopt the RANSAC algorithm, and it can be understood that other fitting algorithms can also be adopted, and the embodiments of the present disclosure are not limited thereto.

[0137] In some embodiments, after fitting at least two pieces of first point cloud data segmented based on the rod point cloud data into first fitted circles, a fitting algorithm such as the RANSAC algorithm can be used to perform linear fitting on the centers of the multiple first fitted circles, and the obtained fitted straight line can be used as the rod center axis of the rod. By this way of first segmenting and fitting to obtain circles, and then fitting a straight line based on the centers of the circles to obtain the rod center axis of the rod, noise caused by irregular shapes such as attachments on the rod and rod branches on the fitting of the rod map element model can be filtered, and good robustness is achieved.

[0138] In some embodiments, after the rod center axis of the rod is determined, a rod map element corresponding to the rod can be modeled based on the rod center axis and the rod point cloud data. In some embodiments, the rod map element can be modeled as a columnar model with the rod center axis as the center line, and the outer surface of the columnar model is modeled based on the collected rod point cloud data, so that a real shape close to the corresponding rod in the real world can be fitted.

[0139] The modeling device for a rod-shaped map element provided in the embodiments of the present disclosure first divides the collected rod-shaped point cloud data into multiple sections of point cloud data in the longitudinal direction; then performs circular fitting on each section of the multiple sections of point cloud data, that is, fitting each section of point cloud data into a circle; then performs straight line fitting on the centers of the multiple fitted circles to obtain the rod center axis of the rod-shaped object, and then models the rod-shaped map element based on the rod center axis and the rod-shaped point cloud data. The modeling device provided in the embodiments of the present disclosure performs segmented circular fitting on the rod-shaped point cloud data, and since the centers of most of the circles obtained through segmented fitting are close to the rod center axis of the real rod-shaped object, the straight line fitting based on these centers can obtain a rod center axis with higher accuracy, and then the segmented modeling is performed on the rod-shaped map element, each section of modeling data corresponds to a circle whose center is located on the rod center axis, and the radius of the circle is determined by the rod-shaped point cloud data corresponding to the section of modeling data, so that the segmented modeling data can obtain the outer surface data close to the real rod-shaped object, thereby obtaining the model data of the rod-shaped map element with high precision and good robustness.

[0140] In an optional implementation of the present embodiment, the first fitting module is specifically configured to:

[0141] project each section of the first point cloud data onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the section of first point cloud data;

[0142] perform circular fitting on the two-dimensional point cloud data to obtain parameters of the first fitted circle corresponding to the first point cloud data, the parameters at least including a center.

[0143] In the optional implementation, as described above, the first point cloud data is a three-dimensional point cloud, and after projecting all or part of the first point cloud data on each section obtained through division onto a plane perpendicular to the longitudinal direction, a set of two-dimensional point cloud located on the plane can be obtained, that is, the three-dimensional point cloud on each section obtained through division is projected onto a plane, and then a circular fitting algorithm is used to perform circular fitting on the obtained two-dimensional point cloud data to obtain a first fitted circle, each section of first point cloud data corresponding to a first fitted circle. After obtaining multiple first fitted circles, the parameters of each first fitted circle can also be obtained, for example, the parameter is a center. In some embodiments, the circular fitting algorithm is, for example, a RANSAC algorithm.

[0144] In some embodiments, the first point cloud data of different sections can be projected onto the same projection plane or different projection planes.

[0145] In an optional implementation of the present embodiment, the first fitting module is specifically configured to:

[0146] set the height of each segment of the first point cloud data as the height of the plane to obtain the corresponding two-dimensional point cloud data.

[0147] In some embodiments, the height of the same segment of the first point cloud data can be set as the height of a plane, such as 0 or other numerical values.

[0148] It can be understood that, in the process of segment fitting, for each segment of the first point cloud data, the density of the point cloud can be greatly increased by compressing the three-dimensional point cloud to a two-dimensional space, thereby overcoming the problem of inaccurate fitting caused by sparse point cloud.

[0149] In the process of segment fitting, the embodiments of the present disclosure compress the first point cloud data of each segment from a three-dimensional point cloud to a two-dimensional space and then perform circular fitting, rather than directly using the first point cloud data of each segment obtained by segmentation to fit a cylinder, thereby reducing the number of fitting parameters and lowering the difficulty of fitting. For the case of a small number of points, the embodiments have better robustness.

[0150] In an optional implementation of the present embodiment, the modeling device further comprises:

[0151] The first determination module is configured to determine the vector distance of each point cloud data in each segment of the first point cloud data to the edge line of the corresponding first fitted circle in the horizontal direction perpendicular to the longitudinal direction;

[0152] The second determination module is configured to determine the fitting residual of the corresponding first fitted circle based on the vector distance of each segment of the first point cloud data.

[0153] The first elimination module is configured to eliminate the first fitted circle whose fitting residual does not satisfy the residual threshold condition from the plurality of first fitted circles.

[0154] In this optional implementation, as described above, the rod-shaped object on the real road is usually not a regular cylinder, and there may be some attachments on it. The center of the first fitted circle obtained by segment fitting may not be on the actual rod center axis of the rod-shaped object. If the actual rod center axis of the rod-shaped object is determined based on the centers of the first fitted circles, a lot of noise may be brought, thereby causing a large deviation between the fitted rod center axis and the actual rod center axis of the rod-shaped object. Therefore, after the first fitted circles are obtained, the present embodiment needs to filter out the noise of the first fitted circles, that is, to eliminate the first fitted circles whose centers are not on the actual rod center axis of the rod-shaped object, and to retain the first fitted circles whose centers are on the actual rod center axis of the rod-shaped object for the next step of straight line fitting.

[0155] In some embodiments, the noise filtering through the fitting residual of the first fitting circle can be performed, and a fitting residual threshold can be determined in advance, and when the fitting residual of the first fitting circle is greater than the determined fitting residual threshold, the first fitting circle is considered to not meet the above requirements and is filtered out as noise.

[0156] In some embodiments, the fitting residual of the first fitting circle can be calculated as follows: in the circle fitting, for the first point cloud data of the same section, each point cloud data in the first point cloud data of the same section is projected on a plane perpendicular to the longitudinal direction to obtain a set of two-dimensional point cloud data, and the circle fitting is performed based on the set of two-dimensional point cloud data to obtain the first fitting circle. In this embodiment, after the first fitting circle is fitted, the vector distance of each two-dimensional point cloud data in the set of two-dimensional point cloud data to the edge line of the first fitting circle can be determined, and the fitting residual of the first fitting circle can be obtained by adding the vector distances corresponding to the set of two-dimensional point cloud data and then dividing by the number of point clouds in the set of two-dimensional point cloud data. It can be understood that the fitting residual of the first fitting circle is not limited to the above calculation method, and the present disclosure does not limit this.

[0157] In some embodiments, the fitting residual threshold can be set to 2 cm, 1 cm or other values, and the present disclosure does not limit this.

[0158] In some embodiments, the fitting residual greater than or equal to the fitting residual threshold can be understood as the fitting residual not meeting the residual threshold condition.

[0159] In this optional implementation, after obtaining the fitting residual of each first fitting circle, it is determined whether the corresponding first fitting circle is qualified by judging whether the fitting residual meets the residual threshold condition; when the fitting residual does not meet the residual threshold condition, it is determined that the first fitting circle is unqualified, so that the first fitting circle can be removed from the plurality of first fitting circles, and the remaining first fitting circles are retained, i.e., the remaining first fitting circles are qualified. This way of removing unqualified first fitting circles through the fitting residual of the first fitting circle can screen out qualified first fitting circles, thereby reducing noise interference and improving fitting accuracy.

[0160] In an optional implementation of the present embodiment, the modeling device further comprises:

[0161] The third determination module is configured to determine a point cloud coverage degree based on the ratio of the length covered by the edge line of the first fitting circle to the circumference of the first fitting circle.

[0162] The second removal module is configured to remove the first fitting circle whose point cloud coverage degree does not meet the coverage degree threshold condition from the plurality of first fitting circles.

[0163] In this optional implementation, as mentioned above, poles on real roads are usually not regular cylinders and may have some attached objects. The center of the first fitted circle obtained by piecewise circle fitting may not be on the actual central axis of the pole. If the actual central axis of the pole is determined by linear fitting based on the centers of these first fitted circles, it may introduce a lot of noise, resulting in a large deviation between the fitted central axis and the actual central axis of the pole. Therefore, in this embodiment, after fitting the first fitted circle, noise filtering is required. That is, the first fitted circles whose centers are likely not on the actual central axis of the pole are discarded, while the first fitted circles whose centers are on the actual central axis of the pole are retained, and the next step of linear fitting is performed.

[0164] In some embodiments, noise is filtered out from the point cloud coverage of the fitted first circle using point cloud data. A coverage threshold can be predetermined. When the point cloud coverage of the first fitted circle is less than or equal to the predetermined coverage threshold, the first fitted circle is considered not to meet the above requirements and is filtered out as noise.

[0165] In some embodiments, the point cloud coverage of the first fitted circle can be understood as the ratio of the length covered by the two-dimensional point cloud data used to fit the first fitted circle along the edge of the first fitted circle to the circumference of the first fitted circle. For example... Figure 2 As shown, the lower middle region illustrates the coverage relationship between a set of two-dimensional point cloud data and a first fitted circle. Assuming the length covered by the two-dimensional point cloud data along the edge of the first fitted circle is L, and the circumference of the first fitted circle is 2πR based on its radius R, the point cloud coverage of the first fitted circle is L / 2πR. In this embodiment, the coverage threshold can be set to 1 / 2, 2 / 3, or other possible values, depending on the actual modeling accuracy requirements. This disclosure does not limit this.

[0166] In some alternative embodiments, the point cloud coverage of the first fitted circle can also be understood as the coverage rate of the two-dimensional point cloud data used to fit the first fitted circle on the circumference of the first fitted circle. For example, the point cloud coverage of the first fitted circle can be determined as the number of degrees the two-dimensional point cloud data covers on the circumference of the first fitted circle. Figure 2 As shown, the coverage relationship between point cloud data and the first fitted circle is illustrated in the lower middle area. In this figure, the point cloud data covers more than half of the circumference of the first fitted circle, that is, the point cloud coverage of the first fitted circle exceeds 180 degrees. In some embodiments, the coverage threshold can be set to 120 degrees, 180 degrees, or other possible degrees, depending on the actual modeling accuracy requirements. This disclosure does not limit this.

[0167] In some embodiments, the point cloud coverage degree is less than or equal to the coverage threshold, which can be understood as that the point cloud coverage degree does not satisfy the coverage threshold condition.

[0168] In this optional implementation, after obtaining the point cloud coverage degree of each first fitting circle, it is determined whether the corresponding first fitting circle is qualified by judging whether the point cloud coverage degree satisfies the coverage threshold; when the point cloud coverage degree does not satisfy the coverage threshold, it can be determined that the first fitting circle is unqualified, so that the first fitting circle is removed from the plurality of first fitting circles, and the remaining first fitting circles are retained, that is, the remaining first fitting circles are qualified. This way of removing unqualified first fitting circles through the fitting residual of the first fitting circle can filter out the first fitting circles that meet the requirements, thereby reducing noise interference and improving fitting accuracy.

[0169] In some embodiments, the first fitting circle can be removed from the plurality of first fitting circles when the fitting residual of the first fitting circle does not satisfy the residual threshold condition and / or the point cloud coverage degree of the first fitting circle does not satisfy the coverage threshold condition, so that the center of the remaining first fitting circle is used for straight line fitting to obtain the rod center axis of the rod-shaped object.

[0170] In an optional implementation of the present embodiment, the modeling module is specifically configured to:

[0171] Based on each piece of first point cloud data, a second fitting circle corresponding to each piece of first point cloud data is fitted, and the center of the second fitting circle is located on the rod center axis;

[0172] The radius and center of the second fitting circle corresponding to the first point cloud data are determined as the modeling data of the rod-shaped map element.

[0173] In the optional implementation, after the rod center axis of the rod-shaped map element is fitted, the modeling data of the rod-shaped map element can be fitted by using the collected first point cloud data and the rod center axis. To this end, in some embodiments, circular fitting can be performed again for each of the at least two segments of first point cloud data obtained by previous segmentation. When the circle is fitted again, the center of the circle can be set on the rod center axis in advance, and a corresponding second fitting circle can be fitted based on a circle fitting algorithm such as the least square method. All the second fitting circles obtained in this way have a point on the rod center axis as the center, and the radius of the circle depends on the distribution of the segment of first point cloud data, which is roughly consistent with the cylindrical surface of the rod-shaped object on the real road. It should be noted that when the second fitting circle is fitted, the intersection of the rod center axis and the projection plane can be taken as the center of the second fitting circle corresponding to the segment of first point cloud data after the segment of first point cloud data is projected onto the same projection plane, and then the radius of the second fitting circle can be fitted based on the center and the segment of first point cloud data. Finally, the second fitting circles corresponding to all the first point cloud data and the radius of each second fitting circle can be taken as the modeling data of the rod-shaped map element, so that an element model consistent with the shape of the corresponding rod-shaped object on the real road can be rendered when the map is rendered.

[0174] In an optional implementation of the present embodiment, the modeling module is specifically configured to:

[0175] segmenting the rod-shaped object point cloud data in the longitudinal direction to obtain at least two segments of second point cloud data;

[0176] fitting, based on each segment of second point cloud data, a third fitting circle corresponding to each segment of second point cloud data, the center of the third fitting circle being located on the rod center axis;

[0177] determining the radius and the center of the third fitting circle corresponding to the second point cloud data as the modeling data of the rod-shaped map element.

[0178] In the optional implementation, after the rod center axis of the rod-shaped map element is fitted, the modeling data of the rod-shaped map element can be fitted by using the collected first point cloud data and the rod center axis. To this end, in some embodiments, the rod-shaped point cloud data can be re-sliced in the longitudinal direction of the rod-shaped point cloud data, and the heights of the multiple pieces of second point cloud data obtained after the slicing can be the same or different. Then, a circular fitting is performed on each piece of second point cloud data in the multiple pieces of second point cloud data obtained after the re-slicing. When the circle is fitted, the center of the circle can be preset on the rod center axis, and a corresponding third fitted circle is fitted based on a circle fitting algorithm such as the least square method. All the third fitted circles obtained in this way have a point on the rod center axis as the center of the circle, and the radius of the circle depends on the distribution of the piece of second point cloud data and is roughly consistent with the cylindrical surface of the rod-shaped object on the real road. It should be noted that when the third fitted circle is fitted, the intersection of the rod center axis and the projection plane after the piece of second point cloud data is projected onto the same projection plane can be taken as the center of the third fitted circle corresponding to the piece of second point cloud data, and then the radius of the third fitted circle is fitted based on the center and the piece of second point cloud data. Finally, the third fitted circles corresponding to all the second point cloud data and the radius of each third fitted circle can be taken as the modeling data of the rod-shaped map element, so that an element model consistent with the shape of the corresponding rod-shaped object on the real road can be rendered when the map is rendered.

[0179] In this embodiment, the rod-shaped point cloud data can be re-sliced based on different requirements for model fitting accuracy. It can be understood that the number of pieces and the heights of the multiple pieces of second point cloud data obtained by re-slicing the rod-shaped point cloud data are different from those of the multiple pieces of first point cloud data obtained by the previous slicing in the above implementation. When the model fitting accuracy requirement is high, the second point cloud data with a lower height can be sliced, that is, compared with the process of slicing the multiple pieces of first point cloud data, the re-slicing of the multiple pieces of second point cloud data has a finer slicing granularity, more third fitted circles are fitted, and the model accuracy of the rod-shaped map element modeled based on the third fitted circles and the rod center axis is higher; while when the model fitting accuracy requirement is low, the second point cloud data with a higher height can be sliced, that is, compared with the process of slicing the multiple pieces of first point cloud data, the re-slicing of the multiple pieces of second point cloud data has a larger slicing granularity, fewer third fitted circles are fitted, and the model accuracy of the rod-shaped map element modeled based on the third fitted circles and the rod center axis is lower, but the calculation efficiency can be improved.

[0180] It should be noted that after the rod center axis of the rod-shaped map element is fitted, whether the first point cloud data obtained by the segmentation is directly used for circular fitting again, or the rod-shaped point cloud data is re-segmented to obtain second point cloud data, and circular fitting is performed based on each segment of the second point cloud data, the modeling data of the rod-shaped map element can be finally obtained. Therefore, by using the segmented circular fitting method, the rod-shaped map element can be represented as a multi-segment cylindrical model sharing a central axis, thereby overcoming the problems of high fitting difficulty and poor fitting accuracy caused by the different thicknesses of the upper and lower parts of the rod-shaped object in the real world.

[0181] The modeling device for the rod-shaped map element provided by the above embodiment can achieve a model fitting accuracy of 2 cm for the vertical rods in the real world, such as the hanging plate rod, street lamp rod, camera rod, and the like in the traffic rod-shaped object, and the success rate is as high as 98%.

[0182] Figure 4 is a structural schematic diagram of an electronic device suitable for implementing the modeling method for the rod-shaped map element according to an embodiment of the present disclosure.

[0183] As shown in Figure 4 , the electronic device 400 includes a processing unit 401, which can be implemented as a CPU, a GPU, an FPGA, an NPU, or the like. The processing unit 401 can perform various processes in the implementation of any of the methods of the present disclosure according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage portion 408 into a random access memory (RAM) 403. Various programs and data required for the operation of the electronic device 400 are also stored in the RAM 403. The processing unit 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0184] The following components are connected to the I / O interface 405: an input portion 406 including a keyboard, a mouse, and the like; an output portion 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker, and the like; a storage portion 408 including a hard disk, and the like; and a communication portion 409 including a network interface card such as a LAN card, a modem, and the like. The communication portion 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is mounted on the drive 410 as needed, so that a computer program read therefrom is installed in the storage portion 408 as needed.

[0185] In particular, according to an embodiment of the present disclosure, the above-mentioned method can be implemented as a computer software program with reference to any one of the embodiments of the present disclosure. For example, an embodiment of the present disclosure includes a computer program product including a computer program tangibly embodied on a machine-readable medium, the computer program containing program code for executing any one of the methods described in the embodiments of the present disclosure. In such an embodiment, the computer program can be downloaded and installed from a network by the communication section 409, and / or installed from the detachable medium 411.

[0186] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowcharts and block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks noted in succession can in fact be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations thereof, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or combinations of hardware and software.

[0187] The units or modules described in the embodiments of the present disclosure can be implemented by software, or by hardware. The described units or modules can also be provided in a processor, and the names of the units or modules do not constitute a limitation on the units or modules themselves in some cases.

[0188] As another aspect, the present disclosure also provides a computer-readable storage medium, which can be the computer-readable storage medium included in the apparatus described in the above embodiments, or can exist separately from the apparatus and not be assembled into the apparatus. The computer-readable storage medium stores one or more programs for execution by one or more processors to perform the methods described in the present disclosure.

[0189] The above description is merely that of the preferred embodiments of the present disclosure and a description of the technical principles of the present disclosure. It should be understood by those skilled in the art that the inventive scope involved in the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by the combinations of the above technical features or equivalent features without departing from the inventive concept. For example, the technical solutions formed by the mutual replacement of the above features and the technical features with similar functions disclosed in the present disclosure (but not limited to) without departing from the inventive concept.

Claims

1. A method of modeling a rod-shaped map element, wherein, The method comprises: acquiring a rod point cloud data; segmenting the rod point cloud data along a longitudinal direction of the rod point cloud data to obtain at least two first point cloud data segments; fitting a first fitting circle corresponding to each first point cloud data segment based on each first point cloud data segment, the parameters of the first fitting circle at least including a center; performing linear fitting on the centers of the first fitting circles to obtain a rod center axis of the rod map element; modeling a rod map element corresponding to the rod based on the rod center axis and the rod point cloud data.

2. The method of claim 1, wherein, The fitting of the first fitting circle corresponding to each first point cloud data segment based on each first point cloud data segment, the parameters of the first fitting circle at least including a center, comprises: projecting each first point cloud data segment onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the first point cloud data segment; performing circular fitting on the two-dimensional point cloud data to obtain the parameters of the first fitting circle corresponding to the first point cloud data, the parameters of the first fitting circle at least including a center.

3. The method of claim 2, wherein, The projection of each first point cloud data segment onto a plane perpendicular to the longitudinal direction to obtain two-dimensional point cloud data corresponding to the first point cloud data segment comprises: setting the height of each first point cloud data segment to the height of the plane to obtain the corresponding two-dimensional point cloud data.

4. The method according to any one of claims 1 to 3, wherein, The parameters further include an edge line of the first fitting circle; after the fitting of the first fitting circle corresponding to each first point cloud data segment based on each first point cloud data segment, the parameters of the first fitting circle at least including a center, the method further comprises: determining the vector distance of each point cloud data in each first point cloud data segment to the edge line of the corresponding first fitting circle in a horizontal direction perpendicular to the longitudinal direction; determining the fitting residual of the corresponding first fitting circle based on the vector distance of each first point cloud data segment; eliminating the first fitting circle whose fitting residual does not satisfy the residual threshold condition from the plurality of first fitting circles.

5. The method of claim 2, wherein, After the fitting of the first fitting circle corresponding to each first point cloud data segment based on each first point cloud data segment, the method further comprises: determining the point cloud coverage degree based on the ratio of the length covered by the edge line of the first fitting circle to the circumference of the first fitting circle corresponding to the two-dimensional point cloud data of the first point cloud data; eliminating the first fitting circle whose point cloud coverage degree does not satisfy the coverage degree threshold condition from the plurality of first fitting circles.

6. The method of any one of claims 1-3, wherein, The modeling of the rod map element corresponding to the rod based on the rod center axis and the rod point cloud data comprises: fitting a second fitting circle corresponding to each first point cloud data segment based on each first point cloud data segment, the center of the second fitting circle being located on the rod center axis; determining the radius and center of the second fitting circle corresponding to the first point cloud data as the modeling data of the rod map element.

7. The method of any one of claims 1-3, wherein, The modeling of the rod map element corresponding to the rod based on the rod center axis and the rod point cloud data comprises: The rod point cloud data is re-segmented in a longitudinal direction of the rod point cloud data to obtain at least two pieces of second point cloud data; Based on each piece of second point cloud data, a third fitting circle corresponding to each piece of second point cloud data is fitted, and a center of the third fitting circle is located on the rod center axis; A radius and a center of the third fitting circle corresponding to the second point cloud data are determined as modeling data of the rod-shaped map element.

8. A device for modeling a rod-shaped map element, wherein, The modeling device comprises: an acquisition module configured to acquire rod point cloud data; a segmentation module configured to segment the rod point cloud data along a longitudinal direction of the rod point cloud data to obtain at least two pieces of first point cloud data; a first fitting module configured to fit a first fitting circle corresponding to each piece of first point cloud data based on each piece of first point cloud data, and the parameters of the first fitting circle at least include a center; a second fitting module configured to perform linear fitting on the center of the first fitting circle to obtain a rod center axis of the rod-shaped map element; a modeling module configured to model a rod-shaped map element corresponding to the rod based on the rod center axis and the rod point cloud data.

9. An electronic device, comprising: A computer program product comprising a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the method of any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon computer instructions, wherein, The computer program product comprises a computer program, and the computer program is executed by a processor to implement the method of any one of claims 1 to 7. The computer program product comprises a computer program, and the computer program is executed by a processor to implement the method of any one of claims 1 to 7.

Images