Real-time positioning map construction method, unmanned mobile device and computer program product

Abstract

The disclosed embodiments disclose a method for constructing a real-time positioning map, an unmanned mobile device, and a computer program product. The method comprises: obtaining point cloud data of an object's surrounding environment; extracting surface features and line features from the point cloud data; performing dimensionality reduction processing on the surface features to obtain reduced-dimensionality line features, and performing dimensionality reduction processing on the line features to obtain reduced-dimensionality point features; and constructing a SLAM map of the object's surrounding environment based on the reduced-dimensionality line features and the reduced-dimensionality point features. This technical solution reduces the computational complexity of the positioning process by performing dimensionality reduction processing on the extracted surface features and line features. Furthermore, because high-dimensional information is utilized during dimensionality reduction to neutralize point cloud noise, the extracted point and line features are more accurate, thereby achieving high-precision, lightweight SLAM technology.

Description

Technical Field

[0001] The present disclosure relates to the field of positioning technology, and in particular to a method for constructing an instant positioning map, an unmanned mobile device, and a computer program product. Background Art

[0002] With the development of technology, unmanned systems such as autonomous vehicles and robotics have become a major research focus at the forefront of technology. Positioning and mapping are paramount in unmanned systems. However, due to limitations in the payload capacity, hardware price, and computing power of unmanned systems, implementing high-performance, real-time positioning and mapping in unmanned systems presents certain technical challenges. Furthermore, because the point cloud data collected by radar is inherently noisy and has a certain degree of randomness, the point cloud has thickness, which, from a visualization perspective, affects the accuracy of feature extraction and matching, thereby affecting the positioning of the unmanned system and the mapping of the surrounding environment.

[0003] Therefore, given the increasing demand for location-based services and the limited load and computing power of unmanned systems (such as autonomous driving and robots), a solution is needed to reduce the amount of computation while ensuring high positioning and mapping accuracy. Summary of the Invention

[0004] The embodiments of the present disclosure provide a method for constructing an instant positioning map, an unmanned mobile device, and a computer program product.

[0005] In a first aspect, an embodiment of the present disclosure provides a method for constructing an instant positioning map, which includes:

[0006] Obtain point cloud data of the object's surrounding environment;

[0007] Extracting surface features and line features from the point cloud data;

[0008] Performing dimensionality reduction processing on the surface features to obtain dimensionality-reduced line features, and performing dimensionality reduction processing on the line features to obtain dimensionality-reduced point features;

[0009] A SLAM map of the surrounding environment of the object is constructed based on the reduced-dimensionality line features and the reduced-dimensionality point features.

[0010] Furthermore, extracting surface features and line features from the point cloud data includes:

[0011] Calculating the curvature of each point in the point cloud data;

[0012] Extracting a face feature based on whether the curvature is less than or equal to a face feature threshold;

[0013] Line features are extracted based on whether the curvature is greater than or equal to a line feature threshold.

[0014] Furthermore, the surface features are subjected to dimensionality reduction processing to obtain dimensionality-reduced line features, including:

[0015] Determining whether two planes formed by the surface feature intersect the line feature based on the spatial position;

[0016] The dimension reduction line feature corresponding to the surface feature is determined based on a straight line formed by the intersection of the two planes.

[0017] Furthermore, determining the dimension reduction line feature corresponding to the surface feature based on the straight line formed by the intersection of the two planes includes:

[0018] Fit the initial plane equations of the two intersecting planes;

[0019] Eliminate the surface features at the intersection of the initial plane equations corresponding to the two planes;

[0020] Optimizing the initial plane equation based on the retained surface features to obtain optimized plane equations of the two planes;

[0021] A dimension-reduced line feature is determined based on an intersection line equation of the two optimized plane equations.

[0022] Furthermore, the line features are subjected to dimensionality reduction processing to obtain dimensionality-reduced point features, including:

[0023] Determining whether there is a surface feature around the line feature based on the spatial position of the line feature;

[0024] When a surface feature exists around the line feature, performing straight line fitting on the line feature to obtain at least two straight line equations;

[0025] The dimension-reduced point features corresponding to the line features are determined based on at least two of the straight line equations.

[0026] Furthermore, determining the dimension reduction point feature corresponding to the line feature based on at least two of the straight line equations includes:

[0027] Eliminate the line features at the intersection of the two straight line equations;

[0028] Optimizing the straight line equation based on the retained line features to obtain two optimized line equations;

[0029] A dimension reduction point feature is determined based on an intersection point of two of the optimization line equations.

[0030] Furthermore, the method further comprises:

[0031] Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the single-frame point cloud data with the constructed SLAM map to obtain a high-frequency, low-precision positioning result of the object;

[0032] Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the sub-map constructed from the multi-frame point cloud data with the SLAM map to obtain a low-frequency and high-precision positioning result of the object;

[0033] Output the high-frequency low-precision positioning result and the low-frequency high-precision positioning result.

[0034] In the second aspect, an embodiment of the present invention provides a location-based service providing method, which includes: obtaining the positioning results of the service object using the method described in the first aspect, and providing location-based services to the service object based on the positioning results, and the location-based services include: one or more of: navigation, route planning, and map rendering.

[0035] In a third aspect, an embodiment of the present invention provides a device for constructing a real-time positioning map, comprising:

[0036] A first acquisition module is configured to acquire point cloud data of the surrounding environment of the object;

[0037] an extraction module configured to extract surface features and line features from the point cloud data;

[0038] a dimensionality reduction module configured to perform dimensionality reduction processing on the surface features to obtain dimensionality-reduced line features, and to perform dimensionality reduction processing on the line features to obtain dimensionality-reduced point features;

[0039] The second acquisition module is configured to construct a SLAM map of the surrounding environment of the object based on the reduced-dimensionality line features and the reduced-dimensionality point features.

[0040] The functions can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the functions.

[0041] In one possible design, the apparatus includes a memory and a processor. The memory is configured to store one or more computer instructions that enable the apparatus to perform the corresponding method, and the processor is configured to execute the computer instructions stored in the memory. The apparatus may also include a communication interface for communicating with other devices or a communication network.

[0042] In a fourth aspect, an embodiment of the present disclosure provides an electronic device, comprising 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 one of the above aspects.

[0043] In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium for storing computer instructions used by any of the above-mentioned devices, and when the computer instructions are executed by a processor, they are used to implement the method described in any of the above-mentioned aspects.

[0044] In a sixth aspect, an embodiment of the present disclosure provides a computer program product, which includes computer instructions, and when the computer instructions are executed by a processor, they are used to implement the method described in any of the above aspects.

[0045] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0046] In the embodiment of the present disclosure, during the construction of a real-time positioning map, an image sensor such as a laser radar mounted on an object is used to scan and obtain three-dimensional point cloud data of the surrounding environment. Surface features and line features are extracted from the three-dimensional point cloud data, and then dimensionality reduction processing is performed on the surface features and line features respectively. That is, the surface features are reduced in dimension to obtain reduced-dimensional line features, and the line features are reduced in dimension to obtain reduced-dimensional point features. The reduced-dimensional line features and reduced-dimensional point features are then used to construct a SLAM map. The embodiment of the present disclosure reduces the amount of computation in the construction of a real-time positioning map by performing dimensionality reduction processing on the extracted surface features and line features. Moreover, because high-dimensional information is used to neutralize point cloud noise during dimensionality reduction, the extracted point and line features are more accurate, thereby achieving a high-precision, lightweight real-time positioning map construction technology.

[0047] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0049] Figure 1 A flowchart showing a method for constructing an instant positioning map according to an embodiment of the present disclosure is shown;

[0050] Figure 2 A schematic diagram of a scenario showing a method for constructing a real-time positioning map according to an embodiment of the present disclosure is shown;

[0051] FIG3( a ) and FIG3 ( b ) show schematic comparison diagrams of two line equations before and after optimization in an embodiment of the present disclosure;

[0052] Figure 4 A schematic diagram illustrating a process of real-time positioning and mapping in an unmanned mobile device according to one embodiment of the present disclosure is shown;

[0053] FIG5( a ) and FIG5 ( b ) are schematic diagrams showing the dimensionality reduction process of surface features and line features, respectively, according to an embodiment of the present disclosure;

[0054] Figure 6 A structural block diagram of a device for constructing a real-time positioning map according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0055] Hereinafter, 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, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0056] In the present disclosure, it should be understood that terms such as "including" or "having" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts, or combinations thereof disclosed in the present specification, and do not exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or combinations thereof exist or are added.

[0057] It should also be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0058] The details of the embodiments of the present disclosure are described in detail below through specific examples.

[0059] Figure 1 FIG. 1 is a flow chart showing a method for constructing a real-time positioning map according to an embodiment of the present disclosure. Figure 1 As shown, the instant positioning map construction method includes the following steps:

[0060] In step S101, point cloud data of the surrounding environment of the object is obtained;

[0061] In step S102, surface features and line features are extracted from the point cloud data;

[0062] In step S103, the surface features are subjected to dimensionality reduction processing to obtain reduced-dimensionality line features, and the line features are subjected to dimensionality reduction processing to obtain reduced-dimensionality point features;

[0063] In step S104, a SLAM map of the surrounding environment of the object is constructed based on the reduced-dimensionality line features and the reduced-dimensionality point features.

[0064] In this embodiment, the real-time positioning map construction method can be executed on an object equipped with an image sensor such as a lidar. The object can be any object, such as an unmanned mobile device such as a drone, an unmanned vehicle, or a robot. The object can be equipped with an image sensor such as a lidar to scan its surroundings while it moves and obtain point cloud data of the surrounding environment. Point cloud data refers to a collection of three-dimensional points with attributes such as coordinates and reflectivity, collected by an image sensor such as a lidar.

[0065] As described in the background technology, simultaneous localization and mapping (SLAM) technology is the key and foundation for objects such as unmanned mobile devices to complete autonomous tracking, detection, navigation and other tasks in unknown environments. Its basic principle is to use various sensors carried by the unmanned mobile device to estimate the position of the unmanned mobile device in the environment. For example, the laser radar on the unmanned mobile device can scan the point cloud image of the surrounding environment. Without considering the world coordinate system, it can construct a local environment map of the surrounding environment based on the scanned point cloud image, and locate its own movement trajectory in the local environment map. The unmanned mobile device can then act according to instructions in the constructed local environment map. If the initial GPS positioning position of the unmanned mobile device when entering the unknown environment is known, the unmanned mobile device can also determine its own GPS trajectory in the world coordinate system based on its own movement trajectory in the local scene map and the initial GPS positioning position.

[0066] Figure 2 A schematic diagram of a scenario of a method for constructing a real-time positioning map according to an embodiment of the present disclosure is shown. Figure 2 As shown, the unmanned mobile device 201 includes a laser radar 2011 and a processing device 2012. After the unmanned mobile device 201 enters an unfamiliar environment, the laser radar 2011 scans the surrounding environment to obtain point cloud data. It can scan 10 times per second, thereby obtaining 10 frames of point cloud data. Each frame of point cloud data obtained by the scan can be processed by the processing device 2012 on the unmanned mobile device 201 to construct a SLAM map of the surrounding environment. The SLAM map is also a real-time positioning map, which can be understood as an environmental map of the environment around the unmanned mobile device constructed using SLAM technology. The mapping process is as follows:

[0067] The processing device 2012 can extract feature points (e.g., points with a large amount of information in the first frame of point cloud data) from the first frame of point cloud data and store the extracted feature points as data in the initial SLAM map. The laser radar 2011 will continue to scan to obtain the second frame of point cloud data, the third frame of point cloud data, and so on. During the scanning process, the unmanned mobile device 201 will continue to move in one or more directions. After the processing device 2012 obtains the second frame of point cloud data from the laser radar, it can extract feature points from the second frame of point cloud data. The extraction method is the same as that in the first frame. The extracted feature points can be matched with the feature points in the initial SLAM map, that is, it is determined which feature points in the initial SLAM map are the same as the feature points extracted in the second frame. Based on the matching results, the processing device 2012 can determine the relative movement information of the second frame of point cloud data relative to the initial SLAM map, such as the movement distance, direction, etc. The processing device 2012 can add the second frame of point cloud data to the initial SLAM map based on the movement information and update the SLAM map. At this time, the SLAM map includes the feature points in the first frame of point cloud data and the second frame of point cloud data. Based on the above method, as time goes by, the laser radar 2011 scans more point cloud data, and the SLAM map constructed by the processing device 2012 also includes more feature points, that is, a more complete surrounding environment map is established.

[0068] The processing device 2012 on the unmanned mobile device 201 can also transmit the above-mentioned SLAM map constructed locally to the back-end server 202, so that the back-end server 202 can provide location services for the unmanned mobile device 201, such as determining the GPS positioning data of the unmanned mobile device 201, and then instructing the movement of the unmanned mobile device 201 based on the GPS positioning data.

[0069] Considering that when building a map locally on an unmanned mobile device, the processing performance of the mobile device is limited, using traditional real-time positioning map building methods will result in low timeliness and accuracy of the results. Therefore, in the disclosed embodiment, after extracting feature points from the point cloud data, the feature points are subjected to dimensionality reduction processing, and then the reduced feature points are used to construct the SLAM map. The implementation details of the disclosed embodiment are described in detail below.

[0070] In this embodiment, an image sensor such as a lidar scans and generates a frame of 3D point cloud data. A processing device on the object extracts surface features and line features from the 3D point cloud data in the current frame. Surface features can be a collection of points located on the same plane, along with information such as their spatial positions. Line features can be points on the same line, such as the edge of a point cloud image, along with information such as their spatial positions.

[0071] In some embodiments, surface features and line features can be determined based on the curvature of each point in the point cloud data. For example, the curvature of each point in the point cloud data can be calculated based on the spatial position relationship between adjacent points. Points with greater curvature than adjacent points are generally edge points and can be determined as line features; points with less curvature are plane points and can be determined as surface features.

[0072] After extracting the surface features and point features from the 3D point cloud data of the current frame, dimensionality reduction processing can be performed on the surface features and line features to reduce the surface features to line features, and the line features to point features. The reduced-dimensional line features and reduced-dimensional point features obtained through dimensionality reduction processing are used as the key features of the current frame to obtain the object positioning result.

[0073] In some embodiments, dimensionality reduction processing of surface features can be understood as reducing the surface features on two intersecting planes into line features on the intersection line of the two intersecting planes, while dimensionality reduction processing of line features can be understood as reducing the line features on two intersecting edge lines into point features at the intersection of the two intersecting edge lines. In this way, the features extracted from the point cloud data can be reduced to features of a lower dimension, which can then be used for matching and positioning.

[0074] In the disclosed embodiment, after extracting the features (including line features and surface features) in the currently scanned point cloud data, the extracted features can be used to match feature points with the SLAM map constructed using the previously scanned point cloud data, and then the SLAM map can be updated based on the matching results, so that the SLAM map includes the features in the currently scanned point cloud data; for example, the positions of certain position points in the surrounding environment in the existing SLAM map and the positions in the currently scanned point cloud data can be determined by matching image features, and then the feature points in the currently scanned point cloud data are added to the constructed SLAM map based on the matched position points, so that the new SLAM map includes images at more locations in the surrounding environment. The details of feature matching in the process of instant positioning map construction can be referred to the existing technology and will not be repeated here.

[0075] In the embodiment of the present disclosure, during the construction of a real-time positioning map, an image sensor such as a laser radar mounted on an object is used to scan and obtain three-dimensional point cloud data of the surrounding environment. Surface features and line features are extracted from the three-dimensional point cloud data, and then dimensionality reduction processing is performed on the surface features and line features respectively. That is, the surface features are reduced in dimension to obtain reduced-dimensional line features, and the line features are reduced in dimension to obtain reduced-dimensional point features. The reduced-dimensional line features and reduced-dimensional point features are then used to construct a SLAM map. The embodiment of the present disclosure reduces the amount of computation in the construction of a real-time positioning map by performing dimensionality reduction processing on the extracted surface features and line features. Moreover, because high-dimensional information is used to neutralize point cloud noise during dimensionality reduction, the extracted point and line features are more accurate, thereby achieving a high-precision, lightweight real-time positioning map construction technology.

[0076] It should be noted that the implementation scenario of the instant positioning map construction in the embodiment of the present disclosure is not limited to Figure 2 The method shown, that is, the process of building a real-time positioning map is not limited to being performed locally on the unmanned mobile device. It can also be performed by scanning point cloud data by image sensors such as lidar on the unmanned mobile device, and the processing equipment on the unmanned mobile device extracts features from the point cloud data, and performs dimensionality reduction processing on the extracted feature points to obtain reduced-dimensionality line features and reduced-dimensionality point features. The reduced-dimensionality line features and reduced-dimensionality point features can be transmitted to the back-end server, and the back-end server performs real-time positioning map construction based on the reduced-dimensionality line features and reduced-dimensionality point features. The specific application scenario can be selected based on actual needs and is not specifically limited here.

[0077] In an optional implementation of this embodiment, step S102, i.e., the step of extracting surface features and line features from the point cloud data, further includes the following steps:

[0078] Calculating the curvature of each point in the point cloud data;

[0079] Extracting a face feature based on whether the curvature is less than or equal to a face feature threshold;

[0080] Line features are extracted based on whether the curvature is greater than or equal to a line feature threshold.

[0081] In this optional implementation, the curvature of each point in the point cloud data can be calculated based on the spatial positional relationship between adjacent points. The curvature calculation can refer to existing technologies and is not specifically limited here. It is understood that points with greater curvature than adjacent points are generally edge points, while points with less curvature are planar points.

[0082] Therefore, in the embodiments of the present disclosure, a surface feature threshold and a line feature threshold can be pre-set, and points with a curvature less than or equal to the surface feature threshold are determined as plane points and then extracted as surface features, while points with a curvature greater than or equal to the line feature threshold are determined as edge points and then extracted as line features. It is understood that the surface feature threshold can be less than or equal to the line feature threshold, and can be set according to actual circumstances, without specific limitation here.

[0083] In some embodiments, the surface features may include features such as the spatial positions of corresponding plane points, and the line features may include features such as the spatial positions of corresponding edge points.

[0084] In an optional implementation of this embodiment, step S103, i.e., the step of performing dimensionality reduction processing on the surface features to obtain reduced-dimensionality line features, further includes the following steps:

[0085] Determining whether two planes formed by the surface feature intersect the line feature based on the spatial position;

[0086] The dimension reduction line feature corresponding to the surface feature is determined based on a straight line formed by the intersection of the two planes.

[0087] In this optional implementation, after extracting the surface features and line features from the point cloud data, dimensionality reduction processing may be performed on the surface features and line features respectively.

[0088] In the process of dimensionality reduction of surface features, it is possible to determine whether there are two planes intersecting with line features based on the surface features. It is understandable that in the point cloud data formed based on three-dimensional objects in the environment, there are usually two intersecting planes. For example, there is a building in the environment, and the front and side of the building are two intersecting planes. Therefore, based on the spatial position relationship of all surface features extracted from the point cloud data, it is possible to determine whether there are two such planes. If so, the features on the intersection line of the two planes can be determined as the reduced-dimensional line features corresponding to the surface features on the two planes, that is, all surface features on the two planes are reduced to line features on the intersection line.

[0089] In some embodiments, when points on a plane are known, a plane equation can be fitted using a conventional mathematical algorithm, such as the RANSAC algorithm. After fitting the plane equations of two intersecting planes using the RANSAC algorithm, the intersection line equation can be calculated based on the two plane equations. Furthermore, the intersection line equation can be used to obtain the dimensionality-reduced line features corresponding to the surface features on the two planes.

[0090] It should be noted that if the point cloud data corresponds to an object with multiple sets of intersecting planes, the dimensionality reduction line features can be calculated for each set of intersecting planes.

[0091] In an optional implementation of this embodiment, the step of determining the dimension-reduced line feature corresponding to the surface feature based on the straight line formed by the intersection of the two planes further includes the following steps:

[0092] Fit the initial plane equations of the two intersecting planes;

[0093] Eliminate the surface features at the intersection of the initial plane equations corresponding to the two planes;

[0094] Optimizing the initial plane equation based on the retained surface features to obtain optimized plane equations of the two planes;

[0095] A dimension-reduced line feature is determined based on an intersection line equation of the two optimized plane equations.

[0096] In this optional implementation, as described above, a mathematical algorithm such as RANSAC can be used to fit the surface features to obtain the plane equations corresponding to the two intersecting planes. Considering that directly fitting the plane equations using the extracted surface features can easily lead to errors in the two fitted planes due to interference from the point cloud data at the intersection of the two planes.

[0097] Therefore, in this embodiment, the two plane equations obtained by fitting the mathematical algorithm are used as the initial plane equations, and the surface features at the intersection of the two planes represented by the two initial plane equations are eliminated, and then the remaining surface features are used to optimize the initial plane equations to obtain two better optimized plane equations. The dimensionality reduction line features are determined based on the intersection line equations of the two optimized plane equations, that is, the intersection line equations are used to represent the dimensionality reduction line features of the surface features on the two intersecting planes.

[0098] In this embodiment, the initial plane equation is optimized based on the principle of minimizing the distance between the point cloud and the plane equation. After removing the surface features at the intersection of the initial plane equation, the disclosed embodiment optimizes the initial plane equation based on the principle of minimizing the distance between the retained surface features and the plane equation. This can offset the errors caused by point cloud noise while preventing interference from the point cloud at the intersection, ultimately obtaining high-precision dimensionality-reduced line features.

[0099] In an optional implementation of this embodiment, step S103, i.e., the step of performing dimensionality reduction processing on the line features to obtain reduced-dimensionality point features, further includes the following steps:

[0100] Determining whether there is a surface feature around the line feature based on the spatial position of the line feature;

[0101] When a surface feature exists around the line feature, performing straight line fitting on the line feature to obtain at least two straight line equations;

[0102] The dimension-reduced point features corresponding to the line features are determined based on at least two of the straight line equations.

[0103] In this optional implementation, after extracting the surface features and line features from the point cloud data, dimensionality reduction processing may be performed on the surface features and line features respectively.

[0104] During the dimensionality reduction process for line features, the presence of surface features around the line feature can be determined based on the spatial position of the line feature. For example, whether the line feature surrounds the surface feature can be determined. If so, it can be determined that there are intersecting edge lines in the line feature. The intersection between the two intersecting edge lines can be determined as the dimensionality reduction point feature corresponding to the two intersecting edge lines.

[0105] In some embodiments, when the points on the line are known, a mathematical algorithm in the prior art can be used to fit the equation of the line. For example, the RANSAC algorithm can be used to fit the equation of the line. For two intersecting edge lines, two line equations can be fitted. If there are multiple sets of intersecting edge lines, more than two line equations can be fitted. After the line equations of the two intersecting edge lines are obtained by fitting using the RANSAC algorithm, the reduced-dimensional point features corresponding to the line features on the two intersecting edge lines can be determined based on the intersection position of the two line equations.

[0106] It should be noted that if the point cloud data corresponds to an object with multiple sets of intersecting edge lines, the dimensionality reduction point features can be calculated for each set of intersecting edge lines.

[0107] In an optional implementation of this embodiment, the step of determining the dimension-reduced point feature corresponding to the line feature based on at least two straight line equations further includes the following steps:

[0108] Eliminate the line features at the intersection of the two straight line equations;

[0109] Optimizing the straight line equation based on the retained line features to obtain two optimized line equations;

[0110] A dimension reduction point feature is determined based on an intersection point of two of the optimization line equations.

[0111] In this optional implementation, as described above, a mathematical algorithm such as RANSAC can be used to fit the line features to obtain the line equations corresponding to the two intersecting edge lines. Considering that directly fitting the line equations using the extracted line features may cause one of the fitted lines to deviate from the other due to interference from the point cloud data at the intersection of the two lines, it is easy to cause errors in the fitted two intersecting edge lines.

[0112] Therefore, in this embodiment, the two straight line equations obtained by fitting the mathematical algorithm are used as the initial straight line equations, and the line features at the intersection of the two lines represented by the two initial straight line equations are eliminated. Then, the initial straight line equations are optimized using the remaining line features retained after eliminating the line features at the above-mentioned intersection, thereby obtaining two better optimized line equations. The dimensionality reduction point features are determined based on the intersection points of the two optimized line equations, that is, the dimensionality reduction point features of the line features on the two intersecting edge lines are represented based on the intersection points of the two optimized line equations.

[0113] In this embodiment, the initial line equation is optimized based on the principle of minimizing the distance from the point cloud to the line equation. After removing the line features at the intersection of the initial line equation, the disclosed embodiment optimizes the initial line equation based on the principle of minimizing the distance from the retained line features to the line equation. This not only offsets the error caused by point cloud noise, but also prevents interference from the point cloud at the intersection from affecting the optimization, ultimately obtaining high-precision dimensionality-reduced point features.

[0114] Figures 3(a) and 3(b) show schematic comparisons of two line equations before and after optimization in an embodiment of the present disclosure. As shown in Figure 3(a), directly fitting two lines using the RANSAC algorithm will cause the line to be pulled toward the other line due to interference from the point cloud at the intersection of the two lines, as shown by the dotted line in Figure 3(a), resulting in inaccurate extracted dimensionality reduction point features.

[0115] Figure 3(b) shows the linear effect of the line equation obtained after removing the point clouds at the intersection of the initially fitted line equation and optimizing the solution using the principle of minimizing the distance from the point to the line. As can be seen from Figure 3(b), the line optimization is accurate after removing the intersecting point clouds, as shown by the dotted line in Figure 3(b).

[0116] In an optional implementation of this embodiment, the method further includes the following steps:

[0117] Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the single-frame point cloud data with the constructed SLAM map to obtain a high-frequency, low-precision positioning result of the object;

[0118] Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the sub-map constructed from the multi-frame point cloud data with the SLAM map to obtain a low-frequency and high-precision positioning result of the object;

[0119] Output the high-frequency low-precision positioning result and the low-frequency high-precision positioning result.

[0120] In this optional implementation, during the positioning of an object, an image sensor such as a laser radar mounted on the object can periodically scan the surrounding environment to obtain point cloud data. A scanning cycle can be a second-level cycle, and the laser radar can scan and obtain a frame of point cloud data within one cycle. Based on the frame of point cloud data, a high-frequency, low-precision positioning result can be obtained. The positioning result can be obtained using the above-mentioned instant positioning map construction method proposed in the embodiment of the present disclosure, that is, extracting surface features and line features from the frame of point cloud data obtained from the current scan, and performing dimensionality reduction processing to obtain corresponding reduced-dimensionality line features and reduced-dimensionality point features. The reduced-dimensionality line features and reduced-dimensionality point features are matched with the constructed SLAM map to obtain a high-frequency, low-precision positioning result of the object. It should be noted that the SLAM map can be a map constructed based on the reduced-dimensionality features corresponding to the point cloud data before the current frame.

[0121] After scanning and obtaining multi-frame point cloud data, the laser radar can also build a sub-map based on the multi-frame point cloud data, extract surface features and line features for the point cloud data in the sub-map, and reduce the dimension of the surface features and line features. The obtained reduced-dimensional line features and reduced-dimensional point features are matched with the SLAM map described above to obtain the low-frequency and high-precision positioning results corresponding to the multi-frame point cloud data.

[0122] The final positioning result of the object can be obtained by fusing the high-frequency low-precision positioning results and the low-frequency high-precision positioning results.

[0123] Figure 4 FIG. 1 is a flow chart showing the process of real-time positioning and mapping in an unmanned mobile device according to an embodiment of the present disclosure. Figure 4 As shown, image sensors on unmanned mobile devices, such as lidar, can scan and acquire point cloud data in real time. After dedistortion, this point cloud data can be extracted using the dimensionality reduction feature extraction method described above to obtain reduced-dimensionality features. Dimensional reduction features extracted from point cloud data obtained by a single-frame scan can be matched with an existing SLAM map at a high frequency of 10Hz, thereby obtaining high-frequency, low-precision poses. Dimensional reduction features extracted from a submap constructed from point cloud data obtained by multiple-frame scans can be matched with the SLAM map at a low frequency of 1Hz, resulting in low-frequency, high-precision poses.

[0124] If the lidar on an unmanned mobile device scans 10 frames of point cloud data per second, the processing device extracts features based on each frame of point cloud data and performs dimensionality reduction. The obtained dimensionality reduction features are matched with the feature points of the constructed SLAM map to obtain the position of the unmanned mobile device in each frame. Based on this position and the dimensionality reduction features extracted from the point cloud data of that frame, the SLAM map can be updated to obtain the positioning result of the unmanned mobile device. In this way, a positioning result can be obtained for each frame of point cloud data, and 10 positioning results can be obtained in 10 frames per second. Because this positioning result can be obtained in a short time, it can be output at a high frequency. However, since each positioning result is obtained based on a single frame of point cloud data, the accuracy is low, and it can be called a high-frequency, low-precision positioning result.

[0125] In addition, in the disclosed embodiment, a positioning result is obtained based on the 10 frames of point cloud data scanned every 1 second. Specifically, the positioning results corresponding to each frame of point cloud data in the 10 frames can be combined for optimization. For example, a graph optimization method is used to optimize the 10 positioning results to obtain a high-precision positioning result. Since this positioning result is only obtained once every 10 frames, that is, once per second, although this positioning result can only be output at a low frequency, it has high accuracy, and is therefore called a low-frequency, high-precision positioning result.

[0126] High-frequency, low-precision positioning results can meet positioning needs with higher timeliness requirements, while low-frequency, high-precision positioning results can meet positioning needs with higher accuracy requirements.

[0127] During the SLAM positioning process, the disclosed embodiment can combine the above-mentioned high-frequency, low-precision positioning results and the low-frequency, high-precision positioning results to output the final positioning results. For example, when scanning 10 frames per second, 10 low-precision positioning results and 1 high-precision positioning result can be output per second, which can meet the needs of both high-frequency positioning applications and low-frequency, high-precision applications.

[0128] According to an embodiment of the present disclosure, an unmanned mobile device includes: an image sensor and a processing device; wherein,

[0129] The image sensor collects point cloud data of the surrounding environment of the unmanned mobile device and outputs the point cloud data to the processing device;

[0130] The processing device extracts surface features and line features from the point cloud data, performs dimensionality reduction processing on the surface features to obtain reduced-dimensionality line features, and performs dimensionality reduction processing on the line features to obtain reduced-dimensionality point features;

[0131] The processing device also constructs a SLAM map of the surrounding environment of the object based on the reduced-dimensionality line features and the reduced-dimensionality point features.

[0132] In this embodiment, the image sensor can be a sensor such as a laser radar installed on the unmanned mobile device that can scan three-dimensional point cloud data of the surrounding environment. The processing device can be a processing unit such as a CPU, GPU, FPGA, or NPU. The processing device can perform the various processes in the above-mentioned SLAM map construction method of the present disclosure according to a program stored in the read-only memory within the unmanned mobile device or a program loaded from an external storage medium into the random access memory.

[0133] In this embodiment, an image sensor scans the surrounding environment of an unmanned mobile device in real time to generate three-dimensional point cloud data. This data is then transmitted to the processing device via a communication link between the image sensor and the processing device. After receiving the three-dimensional point cloud data, the processing device extracts surface features and point features from the data and performs dimensionality reduction on these features to obtain corresponding reduced-dimensionality line features and reduced-dimensionality point features.

[0134] Based on the above-mentioned reduced-dimensionality line features and reduced-dimensionality point features, the processing device can construct a SLAM map of the object's surrounding environment.

[0135] The specific details of the instant positioning map construction method executed by the processing device in the embodiment of the present disclosure can be described above and will not be repeated here.

[0136] According to a location-based service providing method according to an embodiment of the present disclosure, the location-based service providing method includes: using the above-mentioned real-time positioning map construction method to construct a SLAM map of the surrounding environment of the service object and the positioning results in the SLAM map, and providing location-based services to the service object based on the positioning results, wherein the location-based services include: one or more of: navigation, route planning, and map rendering.

[0137] In this embodiment, the location-based service provision method can be executed on a server. The serviced object can be any object equipped with an image sensor for scanning point cloud data in its surrounding environment, such as unmanned systems such as drones, unmanned vehicles, and robots. The image sensor equipped on the serviced object may include, but is not limited to, a lidar, which is used to scan the serviced object's surrounding environment while the serviced object is walking and obtain point cloud data. This point cloud data can be used to obtain the serviced object's positioning result. The process for obtaining the positioning result can be found in the description of the real-time positioning map construction method above and will not be repeated here.

[0138] After extracting feature points from the point cloud data obtained from the current scan, a SLAM map containing the current scan information can be reconstructed based on these feature points and a previously constructed SLAM map. This SLAM map can be a three-dimensional scene map of the object's surrounding environment. Based on the constructed three-dimensional scene map, the location of the service object in the SLAM map and the global positioning result can be obtained. This positioning result can also be used to provide the service object with location services such as navigation, path planning, and map rendering.

[0139] Figures 5(a) and 5(b) respectively show schematic diagrams of the dimensionality reduction processing flow of surface features and line features according to an embodiment of the present disclosure. As shown in Figure 5(a), sensors on unmanned systems, such as lidar, can scan the surrounding environment to obtain point cloud data. After the point cloud data is distorted using a kinematic model, the curvature of all points can be calculated based on the spatial position relationship of adjacent points. Based on the curvature, the ground points in the point cloud data can be segmented to obtain ground points and non-ground points. Subsequent processing processes are all aimed at non-ground points, and ground points can be eliminated. For non-ground points, it can be determined whether their curvature is less than or equal to the surface feature threshold. If it is greater than the surface feature threshold, the point is not a surface feature and is not extracted as a surface feature. If it is less than or equal to the surface feature threshold, the point can be extracted as a surface feature. For all the extracted surface features, it is possible to determine whether there are two planes intersecting the line features based on the spatial position. If not, the surface features do not need to be subjected to dimensionality reduction processing. If they exist, the plane equations can be fitted to these surface features. The RANSAC algorithm can be used for fitting. After the initial surface equation is obtained by fitting, the point cloud data at the intersection of the initial surface equation can be eliminated, and then the initial surface equation can be used as the initial value. The plane equation is optimized and solved based on the principle of minimizing the distance from the point cloud to the plane equation to obtain the optimized surface equation. The line equation obtained by the intersection of the optimized surface equation can be used to represent the dimensionality reduction line features corresponding to these surface features.

[0140] As shown in Figure 5(b), sensors on unmanned systems, such as lidar, can scan the surrounding environment to obtain point cloud data. This point cloud data is then distorted using a kinematic model. The point cloud curvature of all points is calculated based on the spatial relationship between adjacent points. This curvature allows ground points in the point cloud to be segmented into ground and non-ground points. Subsequent processing targets non-ground points, which can be removed. For non-ground points, a determination is made as to whether the point cloud curvature is greater than or equal to a line feature threshold. If the curvature is less than the line feature threshold, the point is considered unlikely to be on an edge line and, therefore, not extracted as a line feature. If the curvature is greater than or equal to the line feature threshold, the point is extracted as a line feature. For the extracted line features, the spatial position of the line features can be used to determine whether they encompass a surface feature. If not, the line features are considered non-intersecting and no dimensionality reduction is performed. If they are, the line features are considered to intersect as edge lines and, therefore, can be reduced to reduced-dimensional point features through dimensionality reduction. Specifically, RANSAC can be used to fit the straight line equations of these line features, remove the point cloud at the intersection of the two straight line equations, use the above fitted straight line equations as the initial value, optimize the execution equation based on the principle of minimizing the distance from the point cloud to the straight line equation, and obtain the optimized line equation. The intersection of the two optimized line equations can be determined as the dimensionality reduction point features corresponding to these line features.

[0141] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0142] Figure 6 The structure block diagram of the instant positioning map construction device according to one embodiment of the present disclosure is shown. The device can be implemented as part or all of an electronic device through software, hardware or a combination of both. Figure 6 As shown, the real-time positioning map building device includes:

[0143] A first acquisition module 601 is configured to acquire point cloud data of the surrounding environment of the object;

[0144] An extraction module 602 is configured to extract surface features and line features from the point cloud data;

[0145] A dimensionality reduction module 603 is configured to perform dimensionality reduction processing on the surface features to obtain reduced-dimensionality line features, and to perform dimensionality reduction processing on the line features to obtain reduced-dimensionality point features;

[0146] The second acquisition module 604 is configured to construct a SLAM map of the surrounding environment of the object based on the reduced-dimensionality line features and the reduced-dimensionality point features.

[0147] In this embodiment, the real-time positioning map construction device can be executed on an object equipped with an image sensor such as a lidar. The object can be any object, such as an unmanned mobile device such as a drone, an unmanned vehicle, or a robot. The object can be equipped with an image sensor such as a lidar to scan its surroundings as it moves and obtain point cloud data of the surrounding environment. Point cloud data refers to a collection of three-dimensional points with attributes such as coordinates and reflectivity, collected by an image sensor such as a lidar.

[0148] In this embodiment, an image sensor such as a lidar scans and generates a frame of 3D point cloud data. A processing device on the object extracts surface features and line features from the 3D point cloud data in the current frame. Surface features can be a collection of points located on the same plane, along with information such as their spatial positions. Line features can be points on the same line, such as the edge of a point cloud image, along with information such as their spatial positions.

[0149] In some embodiments, surface features and line features can be determined based on the curvature of each point in the point cloud data. For example, the curvature of each point in the point cloud data can be calculated based on the spatial position relationship between adjacent points. Points with greater curvature than adjacent points are generally edge points and can be determined as line features; points with less curvature are plane points and can be determined as surface features.

[0150] After extracting the surface features and point features from the 3D point cloud data of the current frame, dimensionality reduction processing can be performed on the surface features and line features to reduce the surface features to line features, and the line features to point features. The reduced-dimensional line features and reduced-dimensional point features obtained through dimensionality reduction processing are used as the key features of the current frame to obtain the object positioning result.

[0151] In some embodiments, dimensionality reduction processing of surface features can be understood as reducing the surface features on two intersecting planes into line features on the intersection line of the two intersecting planes, while dimensionality reduction processing of line features can be understood as reducing the line features on two intersecting edge lines into point features at the intersection of the two intersecting edge lines. In this way, the features extracted from the point cloud data can be reduced to features of a lower dimension, which can then be used for matching and positioning.

[0152] In the disclosed embodiment, after extracting the features (including line features and surface features) in the currently scanned point cloud data, the extracted features can be used to match feature points with the SLAM map constructed using the previously scanned point cloud data, and then the SLAM map can be updated based on the matching results, so that the SLAM map includes the features in the currently scanned point cloud data; for example, the positions of certain position points in the surrounding environment in the existing SLAM map and the positions in the currently scanned point cloud data can be determined by matching image features, and then the feature points in the currently scanned point cloud data are added to the constructed SLAM map based on the matched position points, so that the new SLAM map includes images at more locations in the surrounding environment. The details of feature matching in the process of instant positioning map construction can be referred to the existing technology and will not be repeated here.

[0153] In the embodiment of the present disclosure, during the construction of a real-time positioning map, an image sensor such as a laser radar mounted on an object is used to scan and obtain three-dimensional point cloud data of the surrounding environment. Surface features and line features are extracted from the three-dimensional point cloud data, and then dimensionality reduction processing is performed on the surface features and line features respectively. That is, the surface features are reduced in dimension to obtain reduced-dimensional line features, and the line features are reduced in dimension to obtain reduced-dimensional point features. The reduced-dimensional line features and reduced-dimensional point features are then used to construct a SLAM map. The embodiment of the present disclosure reduces the amount of computation in the construction of a real-time positioning map by performing dimensionality reduction processing on the extracted surface features and line features. Moreover, because high-dimensional information is used to neutralize point cloud noise during dimensionality reduction, the extracted point and line features are more accurate, thereby achieving a high-precision, lightweight real-time positioning map construction technology.

[0154] It should be noted that the implementation scenario of the instant positioning map construction in the embodiment of the present disclosure is not limited to Figure 2 The method shown, that is, the process of building a real-time positioning map is not limited to being performed locally on the unmanned mobile device. It can also be performed by scanning point cloud data by image sensors such as lidar on the unmanned mobile device, and the processing equipment on the unmanned mobile device extracts features from the point cloud data, and performs dimensionality reduction processing on the extracted feature points to obtain reduced-dimensionality line features and reduced-dimensionality point features. The reduced-dimensionality line features and reduced-dimensionality point features can be transmitted to the back-end server, and the back-end server performs real-time positioning map construction based on the reduced-dimensionality line features and reduced-dimensionality point features. The specific application scenario can be selected based on actual needs and is not specifically limited here.

[0155] In an optional implementation of this embodiment, the extraction module includes:

[0156] a calculation submodule, configured to calculate the curvature of each point in the point cloud data;

[0157] A first extraction submodule is configured to extract surface features based on whether the curvature is less than or equal to a surface feature threshold;

[0158] The second extraction submodule is configured to extract line features based on whether the curvature is greater than or equal to a line feature threshold.

[0159] In this optional implementation, the curvature of each point in the point cloud data can be calculated based on the spatial positional relationship between adjacent points. The curvature calculation can refer to existing technologies and is not specifically limited here. It is understood that points with greater curvature than adjacent points are generally edge points, while points with less curvature are planar points.

[0160] Therefore, in the embodiments of the present disclosure, a surface feature threshold and a line feature threshold can be pre-set, and points with a curvature less than or equal to the surface feature threshold are determined as plane points and then extracted as surface features, while points with a curvature greater than or equal to the line feature threshold are determined as edge points and then extracted as line features. It is understood that the surface feature threshold can be less than or equal to the line feature threshold, and can be set according to actual circumstances, without specific limitation here.

[0161] In some embodiments, the surface features may include features such as the spatial positions of corresponding plane points, and the line features may include features such as the spatial positions of corresponding edge points.

[0162] In an optional implementation of this embodiment, the dimensionality reduction module includes:

[0163] A first determining submodule is configured to determine whether two planes formed by the surface feature intersect with the line feature based on spatial position;

[0164] The second determining submodule is configured to determine the dimension reduction line feature corresponding to the surface feature based on the straight line formed by the intersection of the two planes.

[0165] In this optional implementation, after extracting the surface features and line features from the point cloud data, dimensionality reduction processing may be performed on the surface features and line features respectively.

[0166] In the process of dimensionality reduction of surface features, it is possible to determine whether there are two planes intersecting with line features based on the surface features. It is understandable that in the point cloud data formed based on three-dimensional objects in the environment, there are usually two intersecting planes. For example, there is a building in the environment, and the front and side of the building are two intersecting planes. Therefore, based on the spatial position relationship of all surface features extracted from the point cloud data, it is possible to determine whether there are two such planes. If so, the features on the intersection line of the two planes can be determined as the reduced-dimensional line features corresponding to the surface features on the two planes, that is, all surface features on the two planes are reduced to line features on the intersection line.

[0167] In some embodiments, when points on a plane are known, a plane equation can be fitted using a conventional mathematical algorithm, such as the RANSAC algorithm. After fitting the plane equations of two intersecting planes using the RANSAC algorithm, the intersection line equation can be calculated based on the two plane equations. Furthermore, the intersection line equation can be used to obtain the dimensionality-reduced line features corresponding to the surface features on the two planes.

[0168] It should be noted that if the point cloud data corresponds to an object with multiple sets of intersecting planes, the dimensionality reduction line features can be calculated for each set of intersecting planes.

[0169] In an optional implementation of this embodiment, the second determining submodule includes:

[0170] a first fitting submodule, configured to fit an initial plane equation of two intersecting planes;

[0171] A first elimination submodule is configured to eliminate surface features at the intersection of the initial plane equations corresponding to the two planes;

[0172] A first optimization submodule is configured to optimize the initial plane equation based on the retained surface features to obtain optimized plane equations of two planes;

[0173] The third determining submodule is configured to determine the dimension reduction line feature based on the intersection line equation of the two optimized plane equations.

[0174] In this optional implementation, as described above, a mathematical algorithm such as RANSAC can be used to fit the surface features to obtain the plane equations corresponding to the two intersecting planes. Considering that directly fitting the plane equations using the extracted surface features can easily lead to errors in the two fitted planes due to interference from the point cloud data at the intersection of the two planes.

[0175] Therefore, in this embodiment, the two plane equations obtained by fitting the mathematical algorithm are used as the initial plane equations, and the surface features at the intersection of the two planes represented by the two initial plane equations are eliminated, and then the remaining surface features are used to optimize the initial plane equations to obtain two better optimized plane equations. The dimensionality reduction line features are determined based on the intersection line equations of the two optimized plane equations, that is, the intersection line equations are used to represent the dimensionality reduction line features of the surface features on the two intersecting planes.

[0176] In this embodiment, the initial plane equation is optimized based on the principle of minimizing the distance between the point cloud and the plane equation. After removing the surface features at the intersection of the initial plane equation, the disclosed embodiment optimizes the initial plane equation based on the principle of minimizing the distance between the retained surface features and the plane equation. This can offset the errors caused by point cloud noise while preventing interference from the point cloud at the intersection, ultimately obtaining high-precision dimensionality-reduced line features.

[0177] In an optional implementation of this embodiment, the dimensionality reduction module includes:

[0178] A fourth determining submodule is configured to determine whether there is a surface feature around the line feature based on the spatial position of the line feature;

[0179] A second fitting submodule is configured to perform straight line fitting on the line feature to obtain at least two straight line equations when a surface feature exists around the line feature;

[0180] The fifth determining submodule is configured to determine the dimension reduction point feature corresponding to the line feature based on at least two of the straight line equations.

[0181] In this optional implementation, after extracting the surface features and line features from the point cloud data, dimensionality reduction processing may be performed on the surface features and line features respectively.

[0182] During the dimensionality reduction process for line features, the presence of surface features around the line feature can be determined based on the spatial position of the line feature. For example, whether the line feature surrounds the surface feature can be determined. If so, it can be determined that there are intersecting edge lines in the line feature. The intersection between the two intersecting edge lines can be determined as the dimensionality reduction point feature corresponding to the two intersecting edge lines.

[0183] In some embodiments, when the points on the line are known, a mathematical algorithm in the prior art can be used to fit the equation of the line. For example, the RANSAC algorithm can be used to fit the equation of the line. For two intersecting edge lines, two line equations can be fitted. If there are multiple sets of intersecting edge lines, more than two line equations can be fitted. After the line equations of the two intersecting edge lines are obtained by fitting using the RANSAC algorithm, the reduced-dimensional point features corresponding to the line features on the two intersecting edge lines can be determined based on the intersection position of the two line equations.

[0184] It should be noted that if the point cloud data corresponds to an object with multiple sets of intersecting edge lines, the dimensionality reduction point features can be calculated for each set of intersecting edge lines.

[0185] In an optional implementation of this embodiment, the fifth determining submodule includes:

[0186] A second elimination submodule is configured to eliminate line features at the intersection of the two straight line equations;

[0187] a second optimization submodule, configured to optimize the straight line equation based on the retained line features to obtain two optimized line equations;

[0188] The sixth determining submodule is configured to determine a dimension reduction point feature based on an intersection point of two optimized line equations.

[0189] In this optional implementation, as described above, a mathematical algorithm such as RANSAC can be used to fit the line features to obtain the line equations corresponding to the two intersecting edge lines. Considering that directly fitting the line equations using the extracted line features may cause one of the fitted lines to deviate from the other due to interference from the point cloud data at the intersection of the two lines, it is easy to cause errors in the fitted two intersecting edge lines.

[0190] Therefore, in this embodiment, the two straight line equations obtained by fitting the mathematical algorithm are used as the initial straight line equations, and the line features at the intersection of the two lines represented by the two initial straight line equations are eliminated. Then, the initial straight line equations are optimized using the remaining line features retained after eliminating the line features at the above-mentioned intersection, thereby obtaining two better optimized line equations. The dimensionality reduction point features are determined based on the intersection points of the two optimized line equations, that is, the dimensionality reduction point features of the line features on the two intersecting edge lines are represented based on the intersection points of the two optimized line equations.

[0191] In this embodiment, the initial line equation is optimized based on the principle of minimizing the distance from the point cloud to the line equation. After removing the line features at the intersection of the initial line equation, the disclosed embodiment optimizes the initial line equation based on the principle of minimizing the distance from the retained line features to the line equation. This not only offsets the error caused by point cloud noise, but also prevents interference from the point cloud at the intersection from affecting the optimization, ultimately obtaining high-precision dimensionality-reduced point features.

[0192] In an optional implementation of this embodiment, the apparatus includes:

[0193] A first matching module is configured to match the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the single-frame point cloud data with the constructed SLAM map to obtain a high-frequency and low-precision positioning result of the object;

[0194] A second matching module is configured to match the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the sub-map constructed from the multi-frame point cloud data with the SLAM map to obtain a low-frequency and high-precision positioning result of the object;

[0195] The output module is configured to output the high-frequency low-precision positioning result and the low-frequency high-precision positioning result.

[0196] In this optional implementation, during the positioning of an object, the radar scanner mounted on the object can periodically scan the surrounding environment to obtain point cloud data. A scanning cycle can be a second-level cycle, and the laser radar can scan and obtain a frame of point cloud data within one cycle. Based on the frame of point cloud data, a high-frequency, low-precision positioning result can be obtained. The positioning result can be obtained using the above-mentioned instant positioning map construction device proposed in the embodiment of the present disclosure, that is, extracting surface features and line features from the frame of point cloud data obtained from the current scan, and performing dimensionality reduction processing to obtain corresponding reduced-dimensionality line features and reduced-dimensionality point features. The reduced-dimensionality line features and reduced-dimensionality point features are matched with the constructed SLAM map to obtain a high-frequency, low-precision positioning result of the object. It should be noted that the SLAM map can be a map constructed based on the reduced-dimensionality features corresponding to the point cloud data before the current frame.

[0197] After scanning and obtaining multi-frame point cloud data, the laser radar can also build a sub-map based on the multi-frame point cloud data, extract surface features and line features for the point cloud data in the sub-map, and reduce the dimension of the surface features and line features. The obtained reduced-dimensional line features and reduced-dimensional point features are matched with the SLAM map described above to obtain the low-frequency and high-precision positioning results corresponding to the multi-frame point cloud data.

[0198] The final positioning result of the object can be obtained by fusing the high-frequency low-precision positioning results and the low-frequency high-precision positioning results.

[0199] According to an embodiment of the present disclosure, a location-based service providing device includes: using the above-mentioned real-time positioning map construction device to obtain the positioning results of the service object, and providing location-based services to the service object based on the positioning results. The location-based services include: one or more of navigation, route planning, and map rendering.

[0200] In this embodiment, the location-based service provisioning device can be executed on a server, and the serviced object can be any object equipped with a sensor for scanning point cloud data in its surrounding environment, such as unmanned systems such as drones, unmanned vehicles, and robots. Sensors on the serviced object may include, but are not limited to, lidars. These lasers are used to scan the serviced object's surroundings while the serviced object is walking and obtain point cloud data. This point cloud data can be used to obtain the serviced object's positioning results. The process for obtaining positioning results can be found in the description of the real-time positioning map construction device above and will not be further elaborated here.

[0201] After obtaining the location of the service object based on the point cloud data obtained from the current scan, a 3D map containing the current scan information can be constructed based on the location result, the point cloud data, and a previously constructed 3D map. This 3D map can be a 3D map of the object's surroundings. The location result can also be used to provide location services such as navigation and route planning for the service object.

[0202] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the diagram or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, as well as the combination of boxes in the block diagram and / or flow chart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0203] The units or modules described in the embodiments of the present disclosure may be implemented in software or hardware. The units or modules described may also be provided in a processor, and the names of these units or modules do not, in certain circumstances, limit the units or modules themselves.

[0204] As another aspect, the present disclosure further provides a computer-readable storage medium. This computer-readable storage medium may be included in the apparatus described in the above embodiments, or may be a standalone computer-readable storage medium not incorporated into the apparatus. The computer-readable storage medium stores one or more programs, which are used by one or more processors to execute the methods described in the present disclosure.

[0205] The above description is merely a preferred embodiment of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention herein is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

Claims

1. A method for constructing an instant positioning map, wherein: include: Acquiring point cloud data of an environment surrounding an object, wherein the object is an unmanned mobile device; Extracting surface features and line features from the point cloud data; Performing dimensionality reduction processing on the surface features to obtain dimensionality-reduced line features, and performing dimensionality reduction processing on the line features to obtain dimensionality-reduced point features; Constructing a SLAM map of the surrounding environment of the object based on the reduced-dimensionality line features and the reduced-dimensionality point features to reduce the amount of calculation in the process of constructing the real-time positioning map; The dimensionality reduction processing is performed on the surface features to obtain the dimensionality-reduced line features, including: Determining whether two planes formed by the surface feature intersect the line feature based on the spatial position; Fit the initial plane equations of the two intersecting planes; Eliminate the surface features at the intersection of the initial plane equations corresponding to the two planes; Optimizing the initial plane equation based on the retained surface features to obtain optimized plane equations of the two planes; A dimension-reduced line feature is determined based on an intersection line equation of the two optimized plane equations.

2. The method according to claim 1, wherein Extracting surface features and line features from the point cloud data includes: Calculating the curvature of each point in the point cloud data; Extracting a face feature based on whether the curvature is less than or equal to a face feature threshold; Line features are extracted based on whether the curvature is greater than or equal to a line feature threshold.

3. The method according to claim 1 or 2, wherein: Performing dimensionality reduction processing on the line features to obtain dimensionality-reduced point features, including: Determining whether there is a surface feature around the line feature based on the spatial position of the line feature; When a surface feature exists around the line feature, performing straight line fitting on the line feature to obtain at least two straight line equations; The dimension-reduced point features corresponding to the line features are determined based on at least two of the straight line equations.

4. The method according to claim 3, wherein: Determining the dimension-reduced point feature corresponding to the line feature based on at least two of the straight line equations includes: Eliminate the line features at the intersection of the two straight line equations; Optimizing the straight line equation based on the retained line features to obtain two optimized line equations; A dimension reduction point feature is determined based on an intersection point of two of the optimization line equations.

5. The method according to any one of claims 1 to 2 and 4, wherein: The method further comprises: Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the single-frame point cloud data with the constructed SLAM map to obtain a high-frequency, low-precision positioning result of the object; Matching the reduced-dimensionality line features and the reduced-dimensionality point features corresponding to the sub-map constructed from the multi-frame point cloud data with the SLAM map to obtain a low-frequency and high-precision positioning result of the object; Output the high-frequency low-precision positioning result and the low-frequency high-precision positioning result.

6. A method for providing location-based services, wherein: include: The method described in any one of claims 1 to 5 is used to obtain the positioning result of the served object, and based on the positioning result, a location-based service is provided to the served object, wherein the location-based service includes one or more of navigation, route planning, and map rendering.

7. An electronic device, wherein: The method comprises a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 6.

8. A computer program product comprising computer instructions, wherein: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An unmanned mobile device comprising: Image sensor and processor; wherein, The image sensor collects point cloud data of the surrounding environment of the unmanned mobile device and outputs the point cloud data to the processor; The processor extracts surface features and line features from the point cloud data, performs dimensionality reduction processing on the surface features to obtain reduced-dimensionality line features, and performs dimensionality reduction processing on the line features to obtain reduced-dimensionality point features; The processor further constructs a SLAM map of the surrounding environment of the unmanned mobile device based on the reduced dimension line features and the reduced dimension point features to reduce the amount of calculation in the process of constructing the real-time positioning map; The dimensionality reduction processing is performed on the surface features to obtain the dimensionality-reduced line features, including: Determining whether two planes formed by the surface feature intersect the line feature based on the spatial position; Fit the initial plane equations of the two intersecting planes; Eliminate the surface features at the intersection of the initial plane equations corresponding to the two planes; Optimizing the initial plane equation based on the retained surface features to obtain optimized plane equations of the two planes; A dimension-reduced line feature is determined based on an intersection line equation of the two optimized plane equations.

Images