Truth value acquisition method and device, electronic equipment and storage medium

Abstract

Embodiments of the present disclosure disclose a true value acquisition method and device, electronic equipment and storage medium. The method comprises: obtaining existing linear elements based on historical point cloud data and current point cloud data; determining a target point cloud center line matched with the existing linear elements from the current point cloud data; and performing position optimization on the existing linear elements by using the target point cloud center line to obtain a true value linear element. The technical solution can combine the position information in the current point cloud data and the other attribute information of the existing linear elements based on the fact that the position coordinates of the target point cloud center line in the current point cloud data coincide with the linear elements such as lane lines and road edge lines on the real road, and the other attribute information of the existing linear elements is accurate, to obtain the true value linear element, thereby reducing the manual operation time, improving the true value acquisition efficiency, and improving the quality evaluation accuracy and evaluation efficiency of the linear elements identified by the algorithm.

Description

Technical Field

[0001] This disclosure relates to the field of high-precision map technology, specifically to a truth value acquisition method, apparatus, electronic device, and storage medium. Background Technology

[0002] To meet the quality requirements of high-precision map data, it is necessary to evaluate the quality of vector data identified from the latest collected point cloud data, thereby obtaining quality evaluation indicators for the vector data. However, quality evaluation requires corresponding vector ground truth values. Existing technologies obtain vector ground truth values ​​through manual annotation, which is time-consuming, especially for linear elements such as lane lines and road edges, where manual annotation is costly. Furthermore, the vector data to be evaluated is obtained by using algorithms to identify the currently collected point cloud data, while the vector data in existing map data is based on historically collected point cloud data and created manually or through other reliable methods. Due to the continuous changes in roads in the real world, the vector data in existing map data differs from actual roads in the real world and does not meet the ground truth standard, therefore it cannot be used as the ground truth value for vector data identified from the currently collected point cloud data.

[0003] Therefore, a solution is needed to obtain the ground truth of vector data identified from the latest point cloud data at low cost and high efficiency, thereby improving the accuracy and efficiency of quality assessment of the vector data. Summary of the Invention

[0004] This disclosure provides a truth value acquisition method, apparatus, electronic device, and storage medium.

[0005] In a first aspect, this disclosure provides a truth value acquisition method, which includes:

[0006] Acquire existing linear features created based on historical point cloud data, as well as current point cloud data;

[0007] Determine the target point cloud centerline that matches the existing linear features from the current point cloud data;

[0008] The existing linear features are optimized using the centerline of the target point cloud to obtain true linear features.

[0009] Secondly, embodiments of the present invention provide a truth value acquisition device, comprising:

[0010] The first acquisition module is configured to acquire existing linear features created based on historical point cloud data and current point cloud data;

[0011] The determination module is configured to determine, from the current point cloud data, a target point cloud centerline that matches the existing linear feature;

[0012] The second acquisition module is configured to optimize the position of the existing linear features using the center line of the target point cloud to obtain the true linear features.

[0013] The function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above function.

[0014] In one possible design, the above-described device includes a memory and a processor. The memory stores one or more computer instructions that support the device in performing the corresponding methods described above, and the processor is configured to execute the computer instructions stored in the memory. The device may also include a communication interface for communicating with other devices or communication networks.

[0015] Thirdly, embodiments of this disclosure provide an electronic device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method described in any of the above aspects.

[0016] Fourthly, embodiments of this disclosure provide a computer-readable storage medium for storing computer instructions used by any of the above-described devices, which, when executed by a processor, are used to implement the methods described in any of the above aspects.

[0017] Fifthly, embodiments of this disclosure provide a computer program product comprising computer instructions which, when executed by a processor, are used to implement the methods described in any of the preceding aspects.

[0018] The technical solutions provided in this disclosure may have the following beneficial effects:

[0019] In this embodiment, to evaluate the accuracy and other quality of linear features identified by the recognition algorithm from the current point cloud data, ground truth linear features for the aforementioned quality evaluation are obtained based on the current point cloud data and existing linear features. This embodiment first obtains the target point cloud centerline matching the existing linear features based on the current point cloud data. Then, it uses the target point cloud centerline to constrain the position of the existing linear features, optimizing the position of each sampling point on the existing linear features to obtain ground truth linear features. Through this embodiment, based on the fact that the position coordinates of the target point cloud centerline in the current point cloud data match linear features such as lane lines and road edge lines on real roads, and considering the accuracy of other attribute information of existing linear features, the position information in the current point cloud data and other attribute information of existing linear features are combined to obtain ground truth linear features. This reduces manual operation time, improves the efficiency of ground truth acquisition, and thus improves the accuracy and efficiency of quality evaluation of linear features identified by the recognition algorithm.

[0020] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0021] Other features, objects, and advantages of this disclosure will become more apparent from the following detailed description of non-limiting embodiments, taken in conjunction with the accompanying drawings. In the drawings:

[0022] Figure 1 A flowchart illustrating a truth value acquisition method according to an embodiment of the present disclosure is shown.

[0023] Figure 2 A structural block diagram of a truth-finding apparatus according to an embodiment of the present disclosure is shown.

[0024] Figure 3 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0025] Figure 4 This is a schematic diagram of the structure of a computer system suitable for implementing the truth value acquisition method according to an embodiment of the present disclosure. Detailed Implementation

[0026] In the following, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings to enable those skilled in the art to readily implement them. Furthermore, for clarity, portions unrelated to the description of the exemplary embodiments have been omitted from the drawings.

[0027] 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.

[0028] 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.

[0029] The user information (including but not limited to user device information such as location information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals shall be provided for users to choose to authorize or refuse.

[0030] The details of the embodiments of this disclosure are described in detail below through specific examples.

[0031] Figure 1 A flowchart illustrating a truth-value acquisition method according to an embodiment of this disclosure is shown. Figure 1 As shown, the truth value retrieval method includes the following steps:

[0032] In step S101, existing linear features created based on historical point cloud data and current point cloud data are obtained;

[0033] In step S102, a target point cloud centerline that matches the existing linear feature is determined from the current point cloud data;

[0034] In step S103, the existing linear features are optimized using the center line of the target point cloud to obtain true linear features.

[0035] In this embodiment, the truth value acquisition method can be executed on a server. Existing linear features can be linear features in the already created high-precision map data, such as lane lines, road edges, and other continuous, linear map features. Historical point cloud data can be the point cloud data used to create the existing high-precision map data. Current point cloud data can be understood as the most recently collected point cloud data, used to update the existing high-precision map data.

[0036] To ensure the freshness of existing high-precision map data, new point cloud data is typically collected periodically. In this embodiment, the newly collected point cloud data can be referred to as the current point cloud data. To efficiently identify vector data from the current point cloud data, such as linear features in the high-precision map data, a corresponding recognition algorithm can be used to identify the current point cloud data and obtain the latest vector data. This latest vector data can be used to update the existing high-precision map data. To evaluate or improve the recognition accuracy of the algorithm, the quality of the latest vector data identified by the algorithm from the current point cloud data can be evaluated using corresponding ground truth values. However, as described in the background section, the corresponding ground truth values ​​cannot be directly obtained from existing map data. Therefore, this embodiment proposes a ground truth acquisition scheme. This embodiment mainly focuses on obtaining the corresponding ground truth values ​​for linear features.

[0037] To this end, embodiments of this disclosure acquire existing linear features created based on historical point cloud data, i.e., existing linear features in existing high-precision map data, and also acquire current point cloud data. Existing linear features in existing high-precision map data can be, for example, parent database data. It is understood that the current point cloud data can be point cloud data collected for one or more real-world roads, and the existing linear features can be map features already created in existing high-precision map data for the real-world roads covered by the current point cloud data. That is, the area where the existing linear features are located is the same as the area covered by the current point cloud data. It is understood that the current point cloud data can be pre-input into a recognition algorithm, and the recognition algorithm can identify the corresponding vector data, such as lane lines, road edge lines, and other linear features on real-world roads. In some embodiments, the vector data can include, but is not limited to, the position coordinates, length, and direction of each point on the linear feature.

[0038] As described in the background section, in order to evaluate the quality of the linear features identified by the aforementioned recognition algorithm, it is necessary to obtain the corresponding ground truth values. However, due to reasons such as the continuous changes in roads in the real world, the vector data in existing high-precision map data cannot be directly used as the ground truth values ​​of the vector data to be evaluated. To this end, this disclosure proposes a technical solution to optimize existing linear features in existing high-precision map data using current point cloud data, thereby obtaining ground truth linear features that can be used as ground truth values.

[0039] In this embodiment, a target point cloud centerline matching an existing linear feature is extracted from current point cloud data. In some embodiments, semantic segmentation can be performed on the current point cloud data to obtain partial point cloud data that matches the position of an existing linear feature, and then the target point cloud centerline can be extracted from this partial point cloud data. In other embodiments, point cloud lines can also be extracted from the current point cloud data in other ways, and then the target point cloud centerline matching an existing linear feature can be obtained through position matching, etc.

[0040] In some embodiments, the partial point cloud data matching the location of existing linear features can be understood as point cloud data within a region where the existing linear features are located. Processing this partial point cloud data, such as extracting the point cloud center points corresponding to multiple point cloud center points on a real road (e.g., lane lines, road edge lines), can form a target point cloud centerline. In other embodiments, the target point cloud centerline can also be extracted by performing straight line fitting or other methods on the partial point cloud data. It is understood that one or more target point cloud centerlines can be extracted from partial point cloud data. This target point cloud centerline can be understood as corresponding to lane lines, road edge lines, etc., on a real road.

[0041] Considering that real-world roads are constantly changing, there may be errors between the position coordinates of existing linear features and the position coordinates of lane lines, road edge lines, etc., on real roads. This can lead to a distance error between the target point cloud centerline extracted from the current point cloud data and the existing linear feature. The target point cloud centerline corresponds to lane lines, road edge lines, etc., on real roads. When the current point cloud data accuracy is high, the position coordinates of the target point cloud centerline generally match the position coordinates of lane lines, road edge lines, etc., on real roads. Therefore, the position coordinates of ground truth linear features can be referenced to the position coordinates of the target point cloud centerline, while other attribute information of ground truth linear features, such as geometric data and topological connectivity, can be reused from existing linear features. Thus, the position of existing linear features can be constrained based on the target point cloud centerline, and ground truth linear features can be determined based on existing linear features. It is understandable that optimizing the position of an existing linear feature using the target point cloud centerline yields a ground truth linear feature. The other attribute information of the ground truth linear feature is consistent with the corresponding existing linear feature.

[0042] In this embodiment, to evaluate the accuracy and other quality of linear features identified by the recognition algorithm from the current point cloud data, ground truth linear features for the aforementioned quality evaluation are obtained based on the current point cloud data and existing linear features. This embodiment first obtains the target point cloud centerline matching the existing linear features based on the current point cloud data. Then, it uses the target point cloud centerline to constrain the position of the existing linear features, optimizing the position of each sampling point on the existing linear features to obtain ground truth linear features. Through this embodiment, based on the fact that the position coordinates of the target point cloud centerline in the current point cloud data match linear features such as lane lines and road edge lines on real roads, and considering the accuracy of other attribute information of existing linear features, the position information in the current point cloud data and other attribute information of existing linear features are combined to obtain ground truth linear features. This reduces manual operation time, improves the efficiency of ground truth acquisition, and thus improves the accuracy and efficiency of quality evaluation of linear features identified by the recognition algorithm.

[0043] In an optional implementation of this embodiment, step S102, namely the step of determining the target point cloud centerline that matches the existing linear feature from the current point cloud data, further includes the following steps:

[0044] Extract at least one candidate point cloud centerline from the current point cloud data;

[0045] Based on the linear attribute information of the candidate point cloud centerline, a target point cloud centerline that matches the existing linear features is determined.

[0046] In this optional implementation, the current point cloud data can be point cloud data collected from one or more real-world roads, from which one or more candidate point cloud centerlines can be extracted. In some embodiments, the current point cloud data can be semantically segmented, and center points can be extracted based on the segmented partial point cloud data to obtain candidate point cloud centerlines. In other embodiments, candidate point cloud centerlines can also be obtained by performing straight line fitting on the semantically segmented partial point cloud data.

[0047] For the current point cloud data, one or more candidate point cloud centerlines can be obtained. Based on the extracted candidate point cloud centerlines, they can be matched with existing linear features. In some embodiments, the linear attribute information of the candidate point cloud centerlines can be used to match existing linear features. It is understood that there may be multiple existing linear features on the same road, and one or more candidate point cloud centerlines can also be extracted from the current point cloud data collected on the same road. By matching the linear attribute information of the one or more candidate point cloud centerlines with multiple existing linear features, an existing linear feature may match one or more candidate point cloud centerlines, and these one or more candidate point cloud centerlines can be determined as the target point cloud centerlines that match the existing linear feature.

[0048] In some embodiments, the linear attribute information of the candidate point cloud centerline may include, but is not limited to, the position coordinates (including horizontal and vertical coordinates), geometric direction, geometric length, and linear type of the candidate point cloud centerline. In some embodiments, the linear type may include, but is not limited to, solid line and dashed line types. In some embodiments, the linear type may be determined based on the width-to-length ratio of the candidate point cloud centerline. If the width-to-length ratio of the candidate point cloud centerline is large, such as exceeding a preset upper limit, the candidate point cloud centerline can be considered as a dashed line type and can be considered as a segment of a dashed lane line. Conversely, if the width-to-length ratio of the candidate point cloud centerline is small, such as below a preset lower limit, the candidate point cloud centerline can be considered as an implementation type.

[0049] In an optional implementation of this embodiment, the step of determining the target point cloud centerline that matches the existing linear features based on the linear attribute information of the candidate point cloud centerline further includes the following steps:

[0050] The target point cloud centerline is determined based on the horizontal distance, height difference, directional angle, and / or linear type between the candidate point cloud centerline and the existing linear features.

[0051] In this optional implementation, as described above, the position coordinates of existing linear features have a certain error compared to the position coordinates of lane lines, road edge lines, etc. on real roads. Therefore, the position coordinates of the target point cloud centerline can be located around the existing linear feature and within a certain error range. Thus, in this embodiment, a horizontal error range can be predetermined on the horizontal plane, and the horizontal distance from the position coordinates of the candidate point cloud centerline to the position coordinates of the existing linear feature can be used to determine whether the candidate point cloud centerline is within this horizontal error range. Furthermore, the height difference between the two can be used to determine whether the height of the candidate point cloud centerline and the existing linear feature is consistent. If they are inconsistent, it indicates that the candidate point cloud centerline and the existing linear feature belong to different roads, such as one being a ground road and the other an elevated road. The directional angle between the two can also be used to determine whether the geometric direction of the candidate point cloud centerline and the existing linear feature is consistent. Furthermore, the linear type of the two can be determined. If the horizontal distance is within the error range, the height is consistent, the directional angle is less than a preset angle, and the linear type is consistent, then the candidate point cloud centerline can be considered to match the existing linear feature and is the target point cloud centerline corresponding to the existing linear feature.

[0052] In an optional implementation of this embodiment, step S103, which is the step of optimizing the position of the existing linear features using the center line of the target point cloud to obtain the true linear features, further includes the following steps:

[0053] Based at least on the positional error between the centerline of the target point cloud and the existing linear features, the position of the existing linear features is optimized to obtain optimized linear features;

[0054] The confidence level of the optimized linear feature is determined based on its linear attributes and its degree of matching with the center line of the target point cloud.

[0055] The optimized linear elements whose confidence level is higher than a preset confidence threshold are identified as true linear elements.

[0056] In this optional implementation, as mentioned above, the position coordinates of the target point cloud centerline are basically consistent with the position coordinates of lane lines, road edge lines, etc. on the real road. However, the position coordinates of the existing linear elements have a certain error, but other attribute information is accurate. Therefore, based on the position error between the target point cloud centerline and the matching existing linear elements, the position of the existing linear elements can be optimized to obtain optimized linear elements. The difference between the optimized linear elements and the existing linear elements is that the position coordinates of all or some of the position points are different, while other attribute information is the same.

[0057] Considering that some optimized linear features may not meet the standards of true linear features, this embodiment further determines the confidence level of whether an optimized linear feature is a true linear feature based on its own linear attributes and its degree of matching with the center line of the target point cloud. In some embodiments, linear attributes may include, but are not limited to, the smoothness of the optimized linear feature. The degree of matching between the optimized linear feature and the center line of the target point cloud can be measured from multiple aspects, such as the distance error between them, linear attribute error, etc. Linear attributes may include, but are not limited to, linear types, such as solid lines, dashed lines, long lines, short lines, etc. In some embodiments, the confidence level of the optimized linear feature can be obtained by weighted summing of the above-mentioned linear attributes of the optimized linear feature and the degree of matching with the center line of the target point cloud.

[0058] If the smoothness of the optimized linear feature is high, the corresponding index for that linear attribute will be high. Furthermore, if the distance error between the optimized linear feature and the target point cloud centerline is small, then the index for that optimized linear feature can be considered high. If the linearity of the optimized linear feature is consistent with the target point cloud centerline, then the index will also be high. In this case, the confidence level of the optimized linear feature will be high. When this confidence level is higher than a preset confidence threshold, the optimized linear feature can be considered a true linear feature; otherwise, the optimized linear feature may not be a true linear feature and requires further judgment, such as manual judgment.

[0059] In an optional implementation of this embodiment, the step of optimizing the position of the existing linear features based at least on the positional error between the centerline of the target point cloud and the existing linear features to obtain optimized linear features further includes the following steps:

[0060] The position of the original position point on the existing linear feature is optimized by taking the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud as a constraint condition.

[0061] Based on the optimization results, the position coordinates of multiple optimized location points on the optimized linear element are determined.

[0062] In this optional implementation, a distance error equation can be constructed between the target point cloud centerline and existing linear features. This distance error equation is then optimized and solved to minimize the distance error between the optimized positions of each point on the existing linear features and the target point cloud centerline. For example, the sum of the distances from each optimized position on the optimized linear features to the target point cloud centerline can be minimized. When solving this distance error equation, multiple original position points (which can be multiple sampling points) on the existing linear features can be continuously updated to new position points. When the sum of the distances from each of these new position points to the target point cloud centerline is minimized, the optimization result can be considered obtained. This new position point is the optimized position point corresponding to the original position point, thus determining the position coordinates of each optimized position point on the optimized linear feature.

[0063] In an optional implementation of this embodiment, the step of optimizing the position of the original position point on the existing linear feature by taking the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud as a constraint condition further includes the following steps:

[0064] The constraints are the combination of the distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud, the distance error from multiple optimized position points on the optimized linear feature to the original position points on the corresponding existing linear feature, and the smoothness at multiple optimized position points on the optimized linear feature. The original position points on the existing linear feature are then optimized within a preset range.

[0065] In this optional implementation, when optimizing each original position point on an existing linear feature, in addition to using the distance error from multiple optimized position points to the center line of the target point cloud as a constraint, the distance error from multiple optimized position points to the corresponding original position points and the smoothness at multiple optimized position points can also be used as constraints. When these three constraints are combined, the position point with the smallest sum of the three can be the final optimized position point. Furthermore, the position coordinates of the optimized position point cannot exceed the preset range of the original position point.

[0066] The purpose of using the distances from multiple optimized location points to the center line of the target point cloud as constraints is to ensure that the positions of the ground truth linear features closely match the positions of the target point cloud center line, or at least not deviate too much. Therefore, the smaller the distance error between them, the better. Furthermore, the distance between the original location point and the optimized location point should not be too large. Although there is some positional error in the existing linear features, this error should be within a certain range. If the distance between the original location point before optimization and the optimized location point after optimization is too large, it indicates that the optimized location point is not the final optimized location point. Additionally, the optimized location point lies on a ground truth linear feature, which corresponds to lane lines, road edge lines, etc., in the real world and should be a smooth line. If the smoothness at one or more optimized location points is low, there may be broken lines or similar features, therefore, this optimized location point is not the final optimized location point.

[0067] In some embodiments, the original location points on existing linear features can be optimized using the following optimization equation:

[0068]

[0069] Where x is the optimized position point, i is the i-th optimized position point, n is the number of optimized position points, dist is the distance from the i-th optimized position point to the center line of the target point cloud, |δx| is the distance difference between the i-th optimized position point and the original position point, and |gradient| is the gradient at the i-th optimized position point, which is used to represent the smoothness at the i-th optimized position point.

[0070] In an optional implementation of this embodiment, step S102, namely the step of determining the target point cloud centerline that matches the existing linear feature from the current point cloud data, further includes the following steps:

[0071] Based on the position coordinates of the existing linear features, target point cloud data corresponding to the preset range where the existing linear features are located is segmented from the current point cloud data;

[0072] The center line of the target point cloud is determined based on the target point cloud data.

[0073] In this optional implementation, semantic segmentation can be performed on the current point cloud data in advance, that is, according to the position coordinates of the existing linear features, the target point cloud data within the preset range of each existing linear feature is segmented from the current point cloud data.

[0074] Then, the target point cloud centerline can be extracted from the target point cloud data. Since semantic segmentation has been performed on a large area of ​​the current point cloud data in advance, the accuracy of extracting the target point cloud centerline using the segmented target point cloud data can be improved.

[0075] In an optional implementation of this embodiment, the step of determining the centerline of the target point cloud based on the target point cloud data further includes the following steps:

[0076] Extract the center point from the target point cloud data, and determine the center line of the target point cloud based on the center point.

[0077] In this optional implementation, since the target point cloud data is a segment of the current point cloud data based on the position coordinates of existing linear features, and these existing linear features exhibit a linear shape with a large length in the vertical direction and a relatively small width in the horizontal direction, the centerline of the target point cloud can be obtained by extracting the center point in the horizontal direction. That is, by selecting the center point of the target point cloud data in the horizontal direction and connecting the extracted center points in the vertical direction, the centerline of the target point cloud can be obtained.

[0078] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0079] Figure 2 A structural block diagram of a truth-finding apparatus according to an embodiment of the present disclosure is shown. This apparatus can be implemented as part or all of an electronic device through software, hardware, or a combination of both. Figure 2 As shown, the truth-finding device includes:

[0080] The first acquisition module 201 is configured to acquire existing linear features created based on historical point cloud data and current point cloud data;

[0081] The determination module 202 is configured to determine the target point cloud centerline that matches the existing linear feature from the current point cloud data;

[0082] The second acquisition module 203 is configured to optimize the position of the existing linear features using the center line of the target point cloud to obtain the true linear features.

[0083] In this embodiment, the truth-finding device can be executed on a server. Existing linear features can be linear features in the already created high-precision map data, such as lane lines, road edges, and other continuous, linear map features. Historical point cloud data can be the point cloud data used to create the existing high-precision map data. Current point cloud data can be understood as the most recently collected point cloud data, used to update the existing high-precision map data.

[0084] To ensure the freshness of existing high-precision map data, new point cloud data is typically collected periodically. In this embodiment, the newly collected point cloud data can be referred to as the current point cloud data. To efficiently identify vector data from the current point cloud data, such as linear features in the high-precision map data, a corresponding recognition algorithm can be used to identify the current point cloud data and obtain the latest vector data. This latest vector data can be used to update the existing high-precision map data. To evaluate or improve the recognition accuracy of the algorithm, the quality of the latest vector data identified by the algorithm from the current point cloud data can be evaluated using corresponding ground truth values. However, as described in the background section, the corresponding ground truth values ​​cannot be directly obtained from existing map data. Therefore, this embodiment proposes a ground truth acquisition scheme. This embodiment mainly focuses on obtaining the corresponding ground truth values ​​for linear features.

[0085] To this end, embodiments of this disclosure acquire existing linear features created based on historical point cloud data, i.e., existing linear features in existing high-precision map data, and also acquire current point cloud data. Existing linear features in existing high-precision map data can be, for example, parent database data. It is understood that the current point cloud data can be point cloud data collected for one or more real-world roads, and the existing linear features can be map features already created in existing high-precision map data for the real-world roads covered by the current point cloud data. That is, the area where the existing linear features are located is the same as the area covered by the current point cloud data. It is understood that the current point cloud data can be pre-input into a recognition algorithm, and the recognition algorithm can identify the corresponding vector data, such as lane lines, road edge lines, and other linear features on real-world roads. In some embodiments, the vector data can include, but is not limited to, the position coordinates, length, and direction of each point on the linear feature.

[0086] As described in the background art, in order to evaluate the quality of the linear features identified by the above recognition algorithm, it is necessary to obtain the corresponding ground truth. However, due to the fact that roads are constantly changing in the real world, the vector data in the existing high-precision map data cannot be directly used as the ground truth of the vector data to be evaluated. Therefore, the embodiments of this disclosure propose to use the current point cloud data to optimize the existing linear features in the existing high-precision map data to obtain ground truth linear features that can be used as ground truth.

[0087] In this embodiment, a target point cloud centerline matching an existing linear feature is extracted from current point cloud data. In some embodiments, semantic segmentation can be performed on the current point cloud data to obtain partial point cloud data that matches the position of an existing linear feature, and then the target point cloud centerline can be extracted from this partial point cloud data. In other embodiments, point cloud lines can also be extracted from the current point cloud data in other ways, and then the target point cloud centerline matching an existing linear feature can be obtained through position matching, etc.

[0088] In some embodiments, the partial point cloud data matching the location of existing linear features can be understood as point cloud data within a region where the existing linear features are located. Processing this partial point cloud data, such as extracting the point cloud center points corresponding to multiple point cloud center points on a real road (e.g., lane lines, road edge lines), can form a target point cloud centerline. In other embodiments, the target point cloud centerline can also be extracted by performing straight line fitting or other methods on the partial point cloud data. It is understood that one or more target point cloud centerlines can be extracted from partial point cloud data. This target point cloud centerline can be understood as corresponding to lane lines, road edge lines, etc., on a real road.

[0089] Considering that real-world roads are constantly changing, there may be errors between the position coordinates of existing linear features and the position coordinates of lane lines, road edge lines, etc., on real roads. This can lead to a distance error between the target point cloud centerline extracted from the current point cloud data and the existing linear feature. The target point cloud centerline corresponds to lane lines, road edge lines, etc., on real roads. When the current point cloud data accuracy is high, the position coordinates of the target point cloud centerline generally match the position coordinates of lane lines, road edge lines, etc., on real roads. Therefore, the position coordinates of ground truth linear features can be referenced to the position coordinates of the target point cloud centerline, while other attribute information of ground truth linear features, such as geometric data and topological connectivity, can be reused from existing linear features. Thus, the position of existing linear features can be constrained based on the target point cloud centerline, and ground truth linear features can be determined based on existing linear features. It is understandable that optimizing the position of an existing linear feature using the target point cloud centerline yields a ground truth linear feature. The other attribute information of the ground truth linear feature is consistent with the corresponding existing linear feature.

[0090] In this embodiment, to evaluate the accuracy and other quality of linear features identified by the algorithm from the current point cloud data, ground truth linear features for the aforementioned quality evaluation are obtained based on the current point cloud data and existing linear features. This embodiment first obtains the target point cloud centerline matching the existing linear features based on the current point cloud data. Then, it uses the target point cloud centerline to constrain the position of the existing linear features, optimizing the position of each sampling point on the existing linear features to obtain ground truth linear features. Through this embodiment, based on the fact that the position coordinates of the target point cloud centerline in the current point cloud data match linear features such as lane lines and road edge lines on real roads, and considering the accuracy of other attribute information of existing linear features, the position information in the current point cloud data and other attribute information of existing linear features are combined to obtain ground truth linear features. This reduces manual operation time, improves the efficiency of ground truth acquisition, and thus improves the accuracy and efficiency of quality evaluation of the linear features identified by the algorithm.

[0091] In an optional implementation of this embodiment, the determining module may be implemented as follows:

[0092] Extract at least one candidate point cloud centerline from the current point cloud data;

[0093] Based on the linear attribute information of the candidate point cloud centerline, a target point cloud centerline that matches the existing linear features is determined.

[0094] In this optional implementation, the current point cloud data can be point cloud data collected from one or more real-world roads, from which one or more candidate point cloud centerlines can be extracted. In some embodiments, the current point cloud data can be semantically segmented, and center points can be extracted based on the segmented partial point cloud data to obtain candidate point cloud centerlines. In other embodiments, candidate point cloud centerlines can also be obtained by performing straight line fitting on the semantically segmented partial point cloud data.

[0095] For the current point cloud data, one or more candidate point cloud centerlines can be obtained. Based on the extracted candidate point cloud centerlines, they can be matched with existing linear features. In some embodiments, the linear attribute information of the candidate point cloud centerlines can be used to match existing linear features. It is understood that there may be multiple existing linear features on the same road, and one or more candidate point cloud centerlines can also be extracted from the current point cloud data collected on the same road. By matching the linear attribute information of the one or more candidate point cloud centerlines with multiple existing linear features, an existing linear feature may match one or more candidate point cloud centerlines, and these one or more candidate point cloud centerlines can be determined as the target point cloud centerlines that match the existing linear feature.

[0096] In some embodiments, the linear attribute information of the candidate point cloud centerline may include, but is not limited to, the position coordinates (including horizontal and vertical coordinates), geometric direction, geometric length, and linear type of the candidate point cloud centerline. In some embodiments, the linear type may include, but is not limited to, solid line and dashed line types. In some embodiments, the linear type may be determined based on the width-to-length ratio of the candidate point cloud centerline. If the width-to-length ratio of the candidate point cloud centerline is large, such as exceeding a preset upper limit, the candidate point cloud centerline can be considered as a dashed line type and can be considered as a segment of a dashed lane line. Conversely, if the width-to-length ratio of the candidate point cloud centerline is small, such as below a preset lower limit, the candidate point cloud centerline can be considered as an implementation type.

[0097] In an optional implementation of this embodiment, determining the target point cloud centerline that matches the existing linear features based on the linear attribute information of the candidate point cloud centerline can be further implemented as follows:

[0098] The target point cloud centerline is determined based on the horizontal distance, height difference, directional angle, and / or linear type between the candidate point cloud centerline and the existing linear features.

[0099] In this optional implementation, as described above, the position coordinates of existing linear features have a certain error compared to the position coordinates of lane lines, road edge lines, etc. on real roads. Therefore, the position coordinates of the target point cloud centerline can be located around the existing linear feature and within a certain error range. Thus, in this embodiment, a horizontal error range can be predetermined on the horizontal plane, and the horizontal distance from the position coordinates of the candidate point cloud centerline to the position coordinates of the existing linear feature can be used to determine whether the candidate point cloud centerline is within this horizontal error range. Furthermore, the height difference between the two can be used to determine whether the height of the candidate point cloud centerline and the existing linear feature is consistent. If they are inconsistent, it indicates that the candidate point cloud centerline and the existing linear feature belong to different roads, such as one being a ground road and the other an elevated road. The directional angle between the two can also be used to determine whether the geometric direction of the candidate point cloud centerline and the existing linear feature is consistent. Furthermore, the linear type of the two can be determined. If the horizontal distance is within the error range, the height is consistent, the directional angle is less than a preset angle, and the linear type is consistent, then the candidate point cloud centerline can be considered to match the existing linear feature and is the target point cloud centerline corresponding to the existing linear feature.

[0100] In an optional implementation of this embodiment, the second acquisition module may be implemented as:

[0101] Based at least on the positional error between the centerline of the target point cloud and the existing linear features, the position of the existing linear features is optimized to obtain optimized linear features;

[0102] The confidence level of the optimized linear feature is determined based on its linear attributes and its degree of matching with the center line of the target point cloud.

[0103] The optimized linear elements whose confidence level is higher than a preset confidence threshold are identified as true linear elements.

[0104] In this optional implementation, as mentioned above, the position coordinates of the target point cloud centerline are basically consistent with the position coordinates of lane lines, road edge lines, etc. on the real road. However, the position coordinates of the existing linear elements have a certain error, but other attribute information is accurate. Therefore, based on the position error between the target point cloud centerline and the matching existing linear elements, the position of the existing linear elements can be optimized to obtain optimized linear elements. The difference between the optimized linear elements and the existing linear elements is that the position coordinates of all or some of the position points are different, while other attribute information is the same.

[0105] Considering that some optimized linear features may not meet the standards of true linear features, this embodiment further determines the confidence level of whether an optimized linear feature is a true linear feature based on its own linear attributes and its degree of matching with the center line of the target point cloud. In some embodiments, linear attributes may include, but are not limited to, the smoothness of the optimized linear feature. The degree of matching between the optimized linear feature and the center line of the target point cloud can be measured from multiple aspects, such as the distance error between them, linear attribute error, etc. Linear attributes may include, but are not limited to, linear types, such as solid lines, dashed lines, long lines, short lines, etc. In some embodiments, the confidence level of the optimized linear feature can be obtained by weighted summing of the above-mentioned linear attributes of the optimized linear feature and the degree of matching with the center line of the target point cloud.

[0106] If the smoothness of the optimized linear feature is high, the corresponding index for that linear attribute will be high. Furthermore, if the distance error between the optimized linear feature and the target point cloud centerline is small, then the index for that optimized linear feature can be considered high. If the linearity of the optimized linear feature is consistent with the target point cloud centerline, then the index will also be high. In this case, the confidence level of the optimized linear feature will be high. When this confidence level is higher than a preset confidence threshold, the optimized linear feature can be considered a true linear feature; otherwise, the optimized linear feature may not be a true linear feature and requires further judgment, such as manual judgment.

[0107] In an optional implementation of this embodiment, the position of the existing linear features is optimized based at least on the positional error between the centerline of the target point cloud and the existing linear features to obtain optimized linear features. This can be further implemented as follows:

[0108] The position of the original position point on the existing linear feature is optimized by taking the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud as a constraint condition.

[0109] Based on the optimization results, the position coordinates of multiple optimized location points on the optimized linear element are determined.

[0110] In this optional implementation, a distance error equation can be constructed between the target point cloud centerline and existing linear features. This distance error equation is then optimized and solved to minimize the distance error between the optimized positions of each point on the existing linear features and the target point cloud centerline. For example, the sum of the distances from each optimized position on the optimized linear features to the target point cloud centerline can be minimized. When solving this distance error equation, multiple original position points (which can be multiple sampling points) on the existing linear features can be continuously updated to new position points. When the sum of the distances from each of these new position points to the target point cloud centerline is minimized, the optimization result can be considered obtained. This new position point is the optimized position point corresponding to the original position point, thus determining the position coordinates of each optimized position point on the optimized linear feature.

[0111] In an optional implementation of this embodiment, at least the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud is used as a constraint condition to optimize the position of the original position points on the existing linear feature. This can be further implemented as follows:

[0112] The constraints are the combination of the distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud, the distance error from multiple optimized position points on the optimized linear feature to the original position points on the corresponding existing linear feature, and the smoothness at multiple optimized position points on the optimized linear feature. The original position points on the existing linear feature are then optimized within a preset range.

[0113] In this optional implementation, when optimizing each original position point on an existing linear feature, in addition to using the distance error from multiple optimized position points to the center line of the target point cloud as a constraint, the distance error from multiple optimized position points to the corresponding original position points and the smoothness at multiple optimized position points can also be used as constraints. When these three constraints are combined, the position point with the smallest sum of the three can be the final optimized position point. Furthermore, the position coordinates of the optimized position point cannot exceed the preset range of the original position point.

[0114] The purpose of using the distances from multiple optimized location points to the center line of the target point cloud as constraints is to ensure that the positions of the ground truth linear features closely match the positions of the target point cloud center line, or at least not deviate too much. Therefore, the smaller the distance error between them, the better. Furthermore, the distance between the original location point and the optimized location point should not be too large. Although there is some positional error in the existing linear features, this error should be within a certain range. If the distance between the original location point before optimization and the optimized location point after optimization is too large, it indicates that the optimized location point is not the final optimized location point. Additionally, the optimized location point lies on a ground truth linear feature, which corresponds to lane lines, road edge lines, etc., in the real world and should be a smooth line. If the smoothness at one or more optimized location points is low, there may be broken lines or similar features, therefore, this optimized location point is not the final optimized location point.

[0115] In some embodiments, the original location points on existing linear features can be optimized using the following optimization equation:

[0116]

[0117] Where x is the optimized position point, i is the i-th optimized position point, n is the number of optimized position points, dist is the distance from the i-th optimized position point to the center line of the target point cloud, |δx| is the distance difference between the i-th optimized position point and the original position point, and |gradient| is the gradient at the i-th optimized position point, which is used to represent the smoothness at the i-th optimized position point.

[0118] In an optional implementation of this embodiment, the determining module may be implemented as follows:

[0119] Based on the position coordinates of the existing linear features, target point cloud data corresponding to the preset range where the existing linear features are located is segmented from the current point cloud data;

[0120] The center line of the target point cloud is determined based on the target point cloud data.

[0121] In this optional implementation, semantic segmentation can be performed on the current point cloud data in advance, that is, according to the position coordinates of the existing linear features, the target point cloud data within the preset range of each existing linear feature is segmented from the current point cloud data.

[0122] Then, the target point cloud centerline can be extracted from the target point cloud data. Since semantic segmentation has been performed on a large area of ​​the current point cloud data in advance, the accuracy of extracting the target point cloud centerline using the segmented target point cloud data can be improved.

[0123] In an optional implementation of this embodiment, determining the centerline of the target point cloud based on the target point cloud data can be further implemented as follows:

[0124] Extract the center point from the target point cloud data, and determine the center line of the target point cloud based on the center point.

[0125] In this optional implementation, since the target point cloud data is a segment of the current point cloud data based on the position coordinates of existing linear features, and these existing linear features exhibit a linear shape with a large length in the vertical direction and a relatively small width in the horizontal direction, the centerline of the target point cloud can be obtained by extracting the center point in the horizontal direction. That is, by selecting the center point of the target point cloud data in the horizontal direction and connecting the extracted center points in the vertical direction, the centerline of the target point cloud can be obtained.

[0126] This disclosure also discloses an electronic device, Figure 3 This diagram illustrates a structural block diagram of an electronic device according to an embodiment of the present disclosure, such as... Figure 3 As shown, the electronic device 300 includes a memory 301 and a processor 302; wherein,

[0127] The memory 301 is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor 302 to implement the above method steps.

[0128] Figure 4 This is a schematic diagram of the structure of a computer system suitable for implementing the truth-finding method according to an embodiment of the present disclosure.

[0129] like Figure 4 As shown, the computer system 400 includes a processing unit 401, which can be implemented as a CPU, GPU, FPGA, NPU, or other processing units. The processing unit 401 can execute various processes according to any of the methods described above in this disclosure, based on a program stored in the read-only memory (ROM) 402 or a program loaded from the storage portion 408 into the random access memory (RAM) 403. The RAM 403 also stores various programs and data required for the operation of the computer system 400. The processing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0130] The following components are connected to I / O interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to I / O interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 410 as needed so that computer programs read from it can be installed into storage section 408 as needed.

[0131] In particular, according to embodiments of this disclosure, any of the methods described above in the embodiments of this disclosure can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program tangibly embodied on a machine-readable medium, the computer program containing program code for performing any of the methods in the embodiments of this disclosure. In such an embodiment, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411.

[0132] 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 this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0133] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The described units or modules can also be located in a processor, and the names of these units or modules do not necessarily constitute a limitation on the unit or module itself.

[0134] In another aspect, this disclosure also provides a computer-readable storage medium, which may be a computer-readable storage medium included in the apparatus described in the above embodiments; or it may be a standalone computer-readable storage medium not assembled into a device. The computer-readable storage medium stores one or more programs that are used by one or more processors to perform the methods described in this disclosure.

[0135] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for obtaining truth value, wherein, include: Acquire existing linear features created based on historical point cloud data, as well as current point cloud data; Determine the target point cloud centerline that matches the existing linear features from the current point cloud data; Based at least on the positional error between the existing linear features and the target point cloud centerline, the position of the existing linear features is optimized to obtain optimized linear features; based on the linear attributes of the optimized linear features and their matching degree with the target point cloud centerline, the confidence level of the optimized linear features is determined. The optimized linear elements whose confidence level is higher than a preset confidence threshold are identified as true linear elements.

2. The method according to claim 1, wherein, Determining the target point cloud centerline from the current point cloud data that matches the existing linear features includes: Extract at least one candidate point cloud centerline from the current point cloud data; Based on the linear attribute information of the candidate point cloud centerline, a target point cloud centerline that matches the existing linear features is determined.

3. The method according to claim 2, wherein, Based on the linear attribute information of the candidate point cloud centerlines, the target point cloud centerline that matches the existing linear features is determined, including: The target point cloud centerline is determined based on the horizontal distance, height difference, directional angle, and / or linear type between the candidate point cloud centerline and the existing linear features.

4. The method according to claim 1, wherein, Based at least on the positional error between the target point cloud centerline and the existing linear features, the position of the existing linear features is optimized to obtain optimized linear features, including: The position of the original position point on the existing linear feature is optimized by taking the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud as a constraint condition. Based on the optimization results, the position coordinates of multiple optimized location points on the optimized linear element are determined.

5. The method according to claim 4, wherein, The positions of the original points on the existing linear features are optimized by using the minimum distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud as a constraint condition, including: The constraints are the combination of the distance error from multiple optimized position points on the optimized linear feature to the center line of the target point cloud, the distance error from multiple optimized position points on the optimized linear feature to the original position points on the corresponding existing linear feature, and the smoothness at multiple optimized position points on the optimized linear feature. The original position points on the existing linear feature are then optimized within a preset range.

6. The method according to any one of claims 1-2 and 5, wherein, Determining the target point cloud centerline from the current point cloud data that matches the existing linear features includes: Based on the position coordinates of the existing linear features, target point cloud data corresponding to the preset range where the existing linear features are located is segmented from the current point cloud data; The center line of the target point cloud is determined based on the target point cloud data.

7. The method according to claim 6, wherein, The target point cloud centerline is determined based on the target point cloud data, including: Extract the center point from the target point cloud data, and determine the center line of the target point cloud based on the center point.

8. A truth-finding device, wherein, include: The first acquisition module is configured to acquire existing linear features created based on historical point cloud data and current point cloud data; The determination module is configured to determine, from the current point cloud data, a target point cloud centerline that matches the existing linear feature; The second acquisition module is configured to optimize the position of the existing linear features based at least on the positional error between the existing linear features and the center line of the target point cloud, to obtain optimized linear features; determine the confidence level of the optimized linear features based on the linear attributes of the optimized linear features and the degree of matching with the center line of the target point cloud; and determine the optimized linear features with a confidence level higher than a preset confidence threshold as true linear features.

9. An electronic device, wherein, The method includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method of any one of claims 1-7.

10. A computer-readable storage medium having stored thereon computer instructions, wherein, When executed by a processor, the computer instructions implement the method described in any one of claims 1-7.

Images