Semantic map updating method and device, computer device and storage medium

Abstract

The application relates to a semantic map updating method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: extracting a target object point cloud from a crowdsourcing map of a current collection area; constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information; when a semantic map exists in a newly collected collection area, obtaining all second point cloud objects in the semantic map and corresponding second semantic information; matching the crowdsourcing map and the semantic map according to the first semantic information and the second semantic information to obtain a matching result; and updating the semantic map according to the matching result. The method can improve the map updating precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving map and positioning, in particular to a semantic map updating method and device, computer equipment, storage medium and computer program product. BACKGROUND

[0002] With the development of computer technology, automatic driving is the future trend (such as the field of human travel, logistics, etc.). In the solution of automatic driving, in principle, it is the process of realizing the intelligence of the car: "perception-positioning-decision-execution". And high-precision map is the core of perception and positioning. Therefore, the collection, generation and update of the map have also become one of the core technologies of automatic driving.

[0003] At present, there are two common solutions for mapping: professional collection and crowd-sourced collection. The professional collection method is traditional, and generally requires a large number of professional data collection personnel, surveying and mapping equipment, collection vehicles, etc. When collecting high-precision maps, information such as road network data, lane network data, and road traffic facility data needs to be included, so it is often necessary to collect multiple times back and forth on a road to ensure the accuracy of the data. After completion of the collection, it needs to go through data fusion, data processing, release and delivery and many other links. The advantage of this method is that the road data attributes and granularity of the generated high-precision map are very detailed, but the disadvantage is that the update frequency is low, and it is relatively time-consuming and labor-intensive.

[0004] Crowd-sourced collection can be understood as users collecting road data through the sensors of automatic driving vehicles themselves or other low-cost sensors, transmitting the data to the cloud for data fusion, and improving the data accuracy through this fusion method to complete the production of crowd-sourced high-precision maps or semantic maps. At present, most of the crowd-sourced mapping solutions are based on two-dimensional visual data, resulting in sparse and low-precision three-dimensional point clouds, which in turn leads to low precision of automatic driving map updating. SUMMARY

[0005] Therefore, it is necessary to provide a semantic map updating method, device, computer equipment, computer readable storage medium and computer program product capable of improving the precision of map updating in view of the above technical problems.

[0006] In a first aspect, the present application provides a semantic map updating method. The method comprises:

[0007] extracting a target object point cloud from a crowd-sourced map corresponding to a current collection area;

[0008] constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information;

[0009] When the current collection area has a corresponding semantic map, all second point cloud objects in the semantic map and corresponding second semantic information are acquired;

[0010] According to the first semantic information and the second semantic information, the crowd-sourced map is matched with the semantic map to obtain a matching result;

[0011] According to the matching result, the semantic map is updated.

[0012] In one of the embodiments, according to the first semantic information and the second semantic information, the crowd-sourced map is matched with the semantic map to obtain a matching result, including:

[0013] According to the first semantic information and the second semantic information, a semantic distance between the first point cloud object and each second point cloud object is determined;

[0014] According to the semantic distance, the crowd-sourced map is matched with the semantic map to obtain a matching result.

[0015] In one of the embodiments, according to the first semantic information and the second semantic information, a semantic distance between the first point cloud object and each second point cloud object is determined, including:

[0016] According to the first semantic information and the second semantic information, a center position coordinate distance difference value, a point cloud object direction difference, a calibration frame size difference and an appearance feature difference between the first point cloud object and each second point cloud object are determined;

[0017] According to the center position coordinate distance difference value, the point cloud object direction difference value, the calibration frame size difference value and the appearance feature difference value, the semantic distance is determined.

[0018] In one of the embodiments, the first semantic information and the second semantic information include point cloud center coordinates, point cloud calibration frame sizes, point cloud main directions and point cloud histograms, and according to the first semantic information and the second semantic information, a center position coordinate distance difference value, a point cloud object direction difference value, a calibration frame size difference value and an appearance feature difference value between the first point cloud object and each second point cloud object are determined, including:

[0019] According to the point cloud center coordinates of the first point cloud object and the point cloud center coordinates of the second point cloud object, a center position coordinate distance difference value is determined;

[0020] According to the point cloud main direction of the first point cloud object and the point cloud main direction of the second point cloud object, a point cloud object direction difference value is determined;

[0021] determine a bounding box size difference value according to the bounding box size of the first point cloud object and the bounding box size of the second point cloud object;

[0022] determine an appearance feature difference value according to the shape element of the point cloud histogram of the first point cloud object and the shape element of the point cloud histogram of the second point cloud object.

[0023] In one of the embodiments, the determining the semantic distance according to the center position coordinate distance difference value, the point cloud object direction difference value, the bounding box size difference value and the appearance feature difference value comprises:

[0024] weighting the center position coordinate distance difference value, the point cloud object direction difference value, the bounding box size difference value and the appearance feature difference value to obtain the semantic distance between the first point cloud object and the second point cloud object in the semantic map.

[0025] In one of the embodiments, the matching the crowd-sourced map with the semantic map based on the semantic distance to obtain a matching result comprises:

[0026] obtaining the number n of the first point cloud objects in the crowd-sourced map and the number m of the second point cloud objects in the semantic map to obtain an association matrix of n*m;

[0027] taking the semantic distance as an element of the association matrix;

[0028] obtaining the matching result of the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix.

[0029] In one of the embodiments, the obtaining the matching result of the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix comprises:

[0030] when there is a first element in the association matrix greater than or equal to the predetermined threshold, determining that the first point cloud object and the second point cloud object corresponding to the first element are not a point cloud object matching pair.

[0031] In one of the embodiments, the obtaining the matching result of the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix comprises:

[0032] when there is at least one second element in the association matrix less than the predetermined threshold, determining that there is at least one matching second point cloud object for the first point cloud object corresponding to the second element;

[0033] segment the association matrix based on the second elements to obtain a plurality of subgraphs;

[0034] perform bipartite graph matching on each of the subgraphs to determine a second point cloud object matched with the first point cloud object.

[0035] In one of the embodiments, the performing bipartite graph matching on each of the subgraphs to determine a second point cloud object matched with the first point cloud object comprises:

[0036] performing bipartite graph matching on each of the subgraphs to determine a matching value of the first point cloud object and each of the second point cloud objects in the subgraph;

[0037] determining that a second point cloud object corresponding to a minimum matching value is a matching point cloud object of the first point cloud object.

[0038] In one of the embodiments, the updating the semantic map according to the matching result comprises:

[0039] when the matching result is that the first point cloud object is a new point cloud object, adding the first point cloud object to the semantic map.

[0040] In one of the embodiments, the updating the semantic map according to the matching result comprises:

[0041] when the matching result is that the second point cloud object is a disappeared point cloud object, deleting the second point cloud object from the semantic map.

[0042] In one of the embodiments, the updating the semantic map according to the matching result comprises:

[0043] performing fusion processing on the first point cloud object and the matching point cloud object to obtain a fused point cloud object, and determining fused semantic information of the fused point cloud object;

[0044] updating the semantic map according to the fused semantic information.

[0045] In one of the embodiments, the method further comprises:

[0046] when there is no semantic map in the current collection area, creating a semantic map according to the first semantic information.

[0047] In a second aspect, the present application further provides a semantic map updating device. The device comprises:

[0048] an extraction module configured to extract a target point cloud object from a crowdsourcing map corresponding to a current collection area;

[0049] constructing a first point cloud object according to the target object point cloud; the extracting module is further configured to extract first semantic information of the first point cloud object;

[0050] a first obtaining module, configured to, when there is a corresponding semantic map in the current collection area, obtain all second point cloud objects and corresponding second semantic information in the semantic map;

[0051] a matching module, configured to match the crowd-sourced map and the semantic map according to the first semantic information and the second semantic information, to obtain a matching result;

[0052] an updating module, configured to update the semantic map according to the matching result.

[0053] In a third aspect, the present application further provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0054] extracting a target object point cloud from a crowd-sourced map corresponding to a current collection area;

[0055] constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information;

[0056] when there is a corresponding semantic map in the current collection area, obtaining all second point cloud objects and corresponding second semantic information in the semantic map;

[0057] matching the crowd-sourced map and the semantic map according to the first semantic information and the second semantic information, to obtain a matching result;

[0058] updating the semantic map according to the matching result.

[0059] In a fourth aspect, the present application further provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0060] extracting a target object point cloud from a crowd-sourced map corresponding to a current collection area;

[0061] constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information;

[0062] when there is a corresponding semantic map in the current collection area, obtaining all second point cloud objects and corresponding second semantic information in the semantic map;

[0063] match the crowdsourcing map with the semantic map according to the first semantic information and the second semantic information, to obtain a matching result;

[0064] update the semantic map according to the matching result.

[0065] In a fifth aspect, the present application further provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the following steps:

[0066] extract a target object point cloud from a crowdsourcing map corresponding to a current collection area;

[0067] construct a first point cloud object according to the target object point cloud and extract corresponding first semantic information;

[0068] when the current collection area has a corresponding semantic map, obtain all second point cloud objects in the semantic map and corresponding second semantic information;

[0069] match the crowdsourcing map with the semantic map according to the first semantic information and the second semantic information, to obtain a matching result;

[0070] update the semantic map according to the matching result.

[0071] The semantic map updating method, device, computer device, storage medium and computer program product described above extract a target first point cloud object from a crowdsourcing map of a current collection area, and extract first semantic information of the first point cloud object, so that the overall size of the map is greatly reduced on the premise of retaining useful information, and more map information is obtained while the semantic information is extracted and the data size is compressed. When the collection area has a semantic map, the crowdsourcing map is matched with the semantic map according to the second semantic information of all second point cloud objects in the semantic map and the first semantic information, that is, the crowdsourcing collection is matched and updated with the existing semantic map, which provides a guarantee for map matching and updating, and thus improves the map updating accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 An application environment diagram of the semantic map updating method in one embodiment;

[0073] Figure 2 A flowchart of the semantic map updating method in one embodiment;

[0074] Figure 3 A semantic information diagram of a point cloud object in one embodiment;

[0075] Figure 4Fig. 1 is a schematic diagram of partial semantic information of some objects in a local three-dimensional crowdsourcing semantic map in one embodiment;

[0076] Figure 5 Fig. 2 is a schematic diagram of a flow of a method for determining semantic distance of point cloud objects in one embodiment;

[0077] Figure 6 Fig. 3 is a schematic diagram of a flow of a map matching method in one embodiment;

[0078] Figure 7 Fig. 4 is a schematic diagram of a point cloud object matching pair in one embodiment;

[0079] Figure 8 Fig. 5 is a schematic diagram of a flow of a semantic map updating step in one embodiment;

[0080] Figure 9 Fig. 6 is a schematic diagram of a flow of a semantic map updating method in another embodiment;

[0081] Figure 10 Fig. 7 is a schematic diagram of a flow of application of a semantic map updating method in one embodiment;

[0082] Figure 11 Fig. 8 is a structural block diagram of a semantic map updating device in one embodiment;

[0083] Figure 12 Fig. 9 is an internal structure diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0084] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0085] The semantic map updating method provided by the embodiments of the present application can be applied to, for example, Figure 1The application environment shown. Among them, the acquisition device 102 communicates with the terminal 104 through the network. The data storage system can store the data required by the terminal 104 to process. The data storage system can be integrated on the terminal 104, or placed on the cloud or other network servers. The terminal 104 acquires the target object point cloud extracted from the crowdsourcing map corresponding to the current collection area collected by the acquisition device 102; constructs a first point cloud object according to the target object point cloud and extracts the corresponding first semantic information; when there is a semantic map in the newly collected collection area, acquire all second point cloud objects in the semantic map and the corresponding second semantic information; according to the first semantic information and the second semantic information, match the crowdsourcing map with the semantic map to obtain a matching result; update the semantic map according to the matching result. Among them, the terminal 104 can be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices or autonomous driving computing platforms, etc. The acquisition device can be a laser radar, a millimeter wave radar or an ultrasonic radar, or it can be integrated on the terminal. It can be understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and realized through the interaction of the terminal and the server.

[0086] In one embodiment, as shown in Figure 2 , a semantic map updating method is provided. Taking the terminal in Figure 1 as an example, the method includes the following steps:

[0087] Step 202, extracting a target object point cloud from a crowdsourcing map corresponding to a current collection area.

[0088] Among them, crowdsourcing collection can be understood as users collecting road data through the sensors of autonomous vehicles themselves or other low-cost sensors, transmitting the data to the cloud for data fusion, and improving the data accuracy through this fusion method to complete the production of crowdsourcing high-precision maps or semantic maps; The crowdsourcing map or crowdsourcing map refers to the map obtained by the terminal through the cloud, which is collected and uploaded to the cloud by other vehicles. The point cloud data collected by the laser radar contains point clouds that are not needed for constructing a semantic map, and the collected point cloud data needs to be extracted to obtain a target object point cloud.

[0089] The target object point cloud refers to the point cloud of the object required for constructing a semantic map (such as a traffic indication object), and the object required for constructing a semantic map includes but is not limited to traffic lights, traffic signs, lane lines, sidewalks, etc. The target object point cloud can be extracted by existing point cloud target extraction methods, including feature extraction and selection, classification.

[0090] The target object point cloud can be obtained by projecting the 3D point cloud collected by the laser radar into a corresponding 2D image, extracting 2D points in the object calibration frame, and restoring the 2D points to 3D points; or, the point cloud of the object can be obtained by determining the object perception identification point through the perception group of deep learning; wherein, the object calibration frame is identified by the perception deep learning model pre-trained; the perception identification point in the point cloud is also obtained by the deep learning method of the perception group; for example, the 2D traffic light calibration frame is identified by the perception deep learning model; the 3D point cloud of the laser radar is projected into a 2D image, and the points in the calibration frame are extracted and then restored to 3D, which are considered as the points of the traffic light; for another example, the perception identification point of the traffic sign in the point cloud obtained by the deep learning method of the perception group is marked as the point cloud of the traffic sign.

[0091] Specifically, the object required for constructing the semantic map, i.e., the traffic indication object, is determined, and the target object point cloud used to represent the traffic indication object is extracted from the newly collected crowd-sourced map.

[0092] In step 204, a first point cloud object is constructed according to the target object point cloud, and corresponding first semantic information is extracted.

[0093] The semantic information is pre-defined, and each point cloud object has corresponding semantic information, which is different for each point cloud object. The semantic information includes point cloud object identification, point cloud center, point cloud convex hull, point cloud oriented bounding box (OBB) (which can be understood as the minimum calibration frame OBB of the point cloud), point cloud PCA coordinate system direction, point cloud main direction, and point cloud histogram, etc. It can be understood that the target object point cloud here is not limited to one, but can be multiple; and the corresponding first point cloud object can also be one or more.

[0094] Further, the point cloud object identification of each object is different, i.e., unique; the point cloud minimum calibration frame OBB can be represented by eight vertices in the world coordinate system; the point cloud PCA coordinate system direction is calculated by eigenvalue decomposition of the point cloud covariance matrix, and the direction of three dimensions is calculated by the eigenvector (as shown in Figure 3 ), which can be defined as the local coordinate system of the object; the point cloud main direction is the eigenvector corresponding to the smallest eigenvalue after PCA calculation, which is defined as the main direction of the point cloud object. For example, the traffic sign point cloud is approximately a plane, and the normal direction of the plane corresponds to the smallest eigenvalue, which is defined as the main direction (as shown in Figure 3 ); the point cloud convex hull is the smallest convex hull containing all points of an object; and the point cloud histogram is a one-dimensional vector containing the x, y, and z three-dimensional histograms. Figure 4Fig. 2 shows part of the semantic information (including OBB and PCA coordinate system) of some objects in a local three-dimensional crowd-sourced semantic map in one embodiment.

[0095] Specifically, the target object point cloud is clustered and denoised to obtain a series of object point clouds; each object (object class) is constructed through the point cloud, i.e., a first point cloud object corresponding to the point cloud is obtained, and corresponding first semantic information is extracted.

[0096] Step 206, when the current collection area has a corresponding semantic map, all second point cloud objects in the semantic map and corresponding second semantic information are obtained.

[0097] Among them, there are at least one second point cloud object in the semantic map, and each second point cloud object has corresponding second semantic information.

[0098] Specifically, it is determined whether the collection area has a semantic map, and when the newly collected collection area has a semantic map, all second point cloud objects in the semantic map and corresponding second semantic information are obtained; wherein, there are at least one second point cloud object in the semantic map; the second semantic information of the second point cloud object at least includes semantic information including point cloud object identifier, point cloud center, point cloud convex hull, point cloud oriented bounding box (OBB) (which can be understood as the minimum calibration frame OBB of the point cloud), point cloud PCA (Principal Components Analysis) coordinate system direction, point cloud main direction and point cloud histogram, etc. "First" and "second" are only used to distinguish the semantic information of different point cloud objects, for example, the first semantic information can also be named as the second semantic information, and the second semantic information can also be named as the first semantic information.

[0099] Step 208, according to the first semantic information and the second semantic information, the crowd-sourced map and the semantic map are matched to obtain a matching result.

[0100] Among them, the matching result includes point cloud object matching pair and non-point cloud object matching pair; for example, the first point cloud object A has a matching second point cloud object B in the semantic map, then the first point cloud object A and the second point cloud object B are the point cloud object matching pair; the first point cloud object A has no matching second point cloud object in the semantic map, then there is no point cloud object matching pair.

[0101] Specifically, according to the first semantic information and the second semantic information, the semantic distance between the first point cloud object and each second point cloud object is determined; that is, according to the first semantic information and the second semantic information, the center position coordinate distance difference value, the point cloud object direction difference, the calibration box size difference, and the appearance feature difference between the first point cloud object and each second point cloud object are determined; and the semantic distance is determined according to the center position coordinate distance difference value, the point cloud object direction difference value, the calibration box size difference value, and the appearance feature difference value. Based on the semantic distance, the crowd-sourced map is matched with the semantic map to obtain a matching result.

[0102] Further, the point cloud center, the point cloud main direction, the size of the point cloud minimum calibration box, and other physical geometric information in the first semantic information of the first point cloud object in the crowd-sourced map are determined, and the point cloud center, the point cloud main direction, the size of the point cloud minimum calibration box, and other physical geometric information in the second semantic information of the second point cloud object in the semantic map are determined; the semantic distance between the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map is calculated according to the point cloud center, the point cloud main direction, and the size of the point cloud minimum calibration box in the first semantic information and the second semantic information; the obtained semantic distance is weighted with a preset weight value to obtain a final semantic distance; the number n of the first point cloud objects in the crowd-sourced map and the number m of the second point cloud objects in the semantic map are obtained to obtain an association matrix of n*m; the semantic distance is taken as an element of the association matrix; and according to a predetermined threshold value and the association matrix, a matching result of the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map is obtained.

[0103] In one embodiment, the method for determining the semantic distance of the point cloud object includes the following steps:

[0104] In step 502, the center position coordinate distance difference value is determined according to the point cloud center coordinates of the first point cloud object and the point cloud center coordinates of the second point cloud object.

[0105] The first point cloud object is constructed according to the target object point cloud extracted from the newly collected crowd-sourced map, and the first semantic information is the physical geometric information of the point cloud extracted from the first point cloud object, including point cloud object identification, point cloud center, point cloud convex hull, point cloud minimum calibration box OBB, point cloud PCA coordinate system direction, point cloud main direction, and point cloud histogram, and the like.

[0106] Specifically, the target object point cloud is extracted from the crowd-sourced map and a corresponding first point cloud object is constructed, and the first semantic information of the first point cloud object is extracted; the second point cloud object is determined from the existing semantic map, and the second semantic information of the second point cloud object is determined. For example, in the local map (crowd-sourced collection), the first point cloud object object center coordinates are (x1, y1, z1), the point cloud main direction is a, the calibration box size is (w1, h1, d1), the shape feature is a 30-dimensional vector s1, and the original point cloud quantity is n1. On the other hand, in the semantic map, the second point cloud object object center coordinates are (x2, y2, z2), the point cloud main direction is b, the calibration box size is (w2, h2, d2), the shape feature is a 30-dimensional (or other numerical dimension) vector s2, and the original point cloud quantity is n2.

[0107] wherein the center position coordinate difference value is determined according to the point cloud center coordinates (x1, y1, z1) of the first point cloud object and the point cloud center coordinates (x2, y2, z2) of the second point cloud object, and can be expressed as:

[0108] Step 504, the point cloud object direction difference value is determined according to the point cloud main direction of the first point cloud object and the point cloud main direction of the second point cloud object.

[0109] wherein the point cloud object direction difference value can be expressed as:

[0110] cos(θ)=a·b / (|a|·|b|); the greater the angle, the greater the difference, and the smaller the cos value; the smaller the angle, the smaller the difference, and the greater the cos value;

[0111] distance2=1-cos(θ), the smaller the difference, the smaller the direction distance.

[0112] Step 506, the calibration box size difference value is determined according to the calibration box size of the first point cloud object and the calibration box size of the second point cloud object.

[0113] wherein the calibration box size of the first point cloud object is (w1, h1, d1), and the calibration box size of the second point cloud object is (w2, h2, d2); the calibration box size difference value can be expressed as:

[0114]

[0115] Step 508, the appearance feature difference value is determined according to the shape feature of the point cloud histogram of the first point cloud object and the shape feature of the point cloud histogram of the second point cloud object.

[0116] wherein the appearance feature difference value can be expressed as:

[0117]

[0118] Step 510, by weighting the center position coordinate distance difference value, the point cloud object direction difference value, the calibration frame size difference value and the appearance feature difference value, the semantic distance between each first point cloud object and the second point cloud object in the semantic map is obtained.

[0119] Specifically, the weight values w1, w2, w3 and w4 corresponding to the center position coordinate distance difference value distance1, the point cloud object direction difference value distance2, the calibration frame size difference value distance3 and the appearance feature difference value distance4 are obtained; the semantic distance score is obtained by weighted sum of w1, w2, w3 and w4, and the weight value range is 0-1. That is, by adding a weight to each difference value, and then linearly adding, it becomes the final semantic distance; that is, semantic distance = w1*distance1 + w2*distance2 + w3*distance3 + w4*distance4. Alternatively, in an embodiment, according to the difference of actual situation, one of the weight values w1, w2, w3 and w4 can be 0, and two of the weight values can also be 0, which is not limited here.

[0120] Based on the map matching algorithm, according to the determined semantic distance between each first point cloud object and the second point cloud object in the semantic map, the collected crowd-sourced map and the existing semantic map are matched to determine the matching situation between the point cloud objects in the crowd-sourced map and the existing semantic map, wherein the matching situation includes that the first point cloud object in the crowd-sourced map has a matching second point cloud object in the semantic map, the first point cloud object in the crowd-sourced map has no matching second point cloud object in the semantic map (i.e. the first point cloud object is a new point cloud object), and the second point cloud object in the semantic map has no matching first point cloud object in the crowd-sourced map (i.e. the second point cloud object disappears).

[0121] In the above method of determining the semantic distance of the point cloud object, by extracting the semantic information of the first point cloud object in the crowd-sourced map and the semantic information of the second point cloud object in the existing semantic map, the semantic distance between the point cloud object in the crowd-sourced map and the point cloud object in the semantic map is determined according to the object geometric information, which provides a guarantee for map matching update.

[0122] Step 210, updating the semantic map according to the matching result.

[0123] The matching method is Hungarian matching, but is not limited thereto. The matching result includes two cases: a matching pair of objects in the crowdsourcing map and the semantic map, and a non-matching pair of objects in the crowdsourcing map and the semantic map. Further, the matching pair of objects in the crowdsourcing map and the semantic map means that the first point cloud object in the crowdsourcing map has a matching second point cloud object in the semantic map. The non-matching pair of objects in the crowdsourcing map and the semantic map includes two cases: the first point cloud object in the crowdsourcing map has no corresponding matching point cloud object in the semantic map, and the second point cloud object in the semantic map has no corresponding matching point cloud object in the crowdsourcing map. In this patent, for the convenience of understanding, the first point cloud object in the crowdsourcing map that is not matched in the first case is referred to as a new point cloud object (i.e., the first point cloud object is a newly added object), and the second point cloud object in the semantic map that is not matched in the second case is referred to as a disappeared point cloud object (i.e., the second point cloud object does not exist in the newly collected crowdsourcing map).

[0124] Specifically, when the matching result is that the first point cloud object in the crowdsourcing map has a matching second point cloud object in the semantic map, the semantic information of the first point cloud object and the second point cloud object is processed by weighted average to obtain the semantic information of a new point cloud object; the semantic map is updated according to the semantic information of the new point cloud object; when the first point cloud object in the crowdsourcing map is a newly added object, the first point cloud object is added in the semantic map; and when the second point cloud object does not exist in the newly collected crowdsourcing map, the original second point cloud object in the semantic map is deleted to obtain an updated semantic map.

[0125] In the above semantic map updating method, the target first point cloud object is extracted from the newly collected crowdsourcing map, and the first semantic information of the first point cloud object is extracted, i.e., the semantic information is extracted on the premise of retaining useful information, so that the updating data required in the map updating process is greatly reduced, more map information is obtained while the data size is compressed, when there is no semantic map in the collection area, the newly collected crowdsourcing map can be spliced with the semantic map to update the entire map; when there is a semantic map in the collection area, only the original semantic map can be incrementally updated, the crowdsourcing map is matched with the semantic map according to the second semantic information of all second point cloud objects in the semantic map and the first semantic information, i.e., the newly collected data is matched and updated with the existing semantic map, which provides a guarantee for the map matching and updating, and further improves the map updating accuracy.

[0126] In one embodiment, as shown in FIG. 1, Figure 6 a map matching method is provided. The method is applied to a terminal in Figure 1 for example, and includes the following steps:

[0127] Step 602, extracting a target object point cloud from a crowdsourcing map corresponding to a current acquisition area.

[0128] Step 604, constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information.

[0129] Specifically, the extracted target object point cloud is denoised and clustered to construct at least one first point cloud object, and first semantic information of each first point cloud object is extracted.

[0130] Step 606, obtaining all second point cloud objects in a semantic map and corresponding second semantic information.

[0131] Step 608, determining semantic distances between the first point cloud object and each second point cloud object according to the first semantic information and the second semantic information.

[0132] Specifically, a center position coordinate distance difference value is determined according to a point cloud center coordinate of the first point cloud object and a point cloud center coordinate of the second point cloud object; a point cloud object direction difference value is determined according to a point cloud main direction of the first point cloud object and a point cloud main direction of the second point cloud object; a calibration frame size difference value is determined according to a calibration frame size of the first point cloud object and a calibration frame size of the second point cloud object; an appearance feature difference value is determined according to a shape element of a point cloud histogram of the first point cloud object and a shape element of a point cloud histogram of the second point cloud object; and a semantic distance between each first point cloud object and a second point cloud object in the semantic map is obtained by weighting a center position coordinate distance difference score, a point cloud object direction difference score, a calibration frame size difference score and an appearance feature difference score.

[0133] Step 610, obtaining a number n of the first point cloud objects in the crowdsourcing map and a number m of the second point cloud objects in the semantic map to obtain an association matrix of n*m.

[0134] Step 612, taking the semantic distance as an element of the association matrix.

[0135] It can be understood that each element in the association matrix represents a semantic distance between a first point cloud object in the crowdsourcing map and a second point cloud object in the semantic map; for example, the number of first point cloud objects in the crowdsourcing map and the number of second point cloud objects in the semantic map are 3, forming a 3*3 association matrix, and each element in the association matrix is a semantic distance between a first point cloud object and a second point cloud object in the semantic map.

[0136] Step 614, obtaining a matching result of the first point cloud object in the crowdsourcing map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix.

[0137] Wherein, the semantic distance less than the predetermined threshold value indicates that the corresponding first point cloud object and the second point cloud object exist correlation.

[0138] Specifically, the semantic distance is taken as an element of the correlation matrix, when there is a first element greater than or equal to the predetermined threshold value in the correlation matrix, it is determined that the first point cloud object and the second point cloud object corresponding to the first element are not point cloud object matching pairs; the first point cloud object and the second point cloud object corresponding to the first element are not point cloud object matching pairs, including that the first point cloud object is a newly added point cloud object or the second point cloud object is a disappeared point cloud object.

[0139] When there is at least one second element less than the predetermined threshold value in the correlation matrix, it is determined that the first point cloud object corresponding to the second element has at least one matched second point cloud object; the correlation matrix is segmented based on the second element to obtain a plurality of subgraphs; bipartite graph matching is performed on each subgraph to determine the second point cloud object matched with the first point cloud object; that is, bipartite graph matching is performed on each subgraph to determine the matching cost value of the first point cloud object and each second point cloud object in the subgraph; the second point cloud object corresponding to the minimum matching cost value is determined as the matching point cloud object of the first point cloud object. Wherein, the bipartite graph matching is realized by using the Hungarian algorithm, and the object connection pair (object, crowd_object) with the minimum cost is obtained by using the Hungarian algorithm for bipartite graph matching, as shown in Figure 7 The matching pair of point clouds may be obtained; in other words, the correlation degree (the correlation degree can be understood as the matching cost value) of each second point cloud object in the subgraph and the corresponding first point cloud object is obtained.

[0140] For example, the crowd-sourced collected map includes first point cloud objects 1, 2, 3, and the existing semantic map has second point cloud objects 4, 5, 6, wherein the semantic distance between the first point cloud object 2 and the second point cloud object 4 and the first point cloud object 2 and the second point cloud object 5 is greater than the predetermined value, which is a non-matching point cloud object, the semantic distance between the first point cloud object 1 and the second point cloud object 4 and the first point cloud object 1 and the second point cloud object 5 is less than the predetermined threshold value, then the semantic distance between the first point cloud object 1 and the second point cloud object 4 and the first point cloud object 1 and the second point cloud object 5 needs to be segmented into a subgraph, and the existing Hungarian matching algorithm is used for the first point cloud object 1 and the second point cloud object 4 and the first point cloud object 1 and the second point cloud object 5 to obtain the matching cost value cost, and the minimum cost is determined as the final matching point cloud object of the first point cloud object 1.

[0141] In the above map matching method, the target object point cloud is extracted from the crowd-sourced data, and the semantic information is further extracted based on the extracted object point cloud, so that more information is obtained while the data size is compressed; the entire map data size is extremely small, and all the required information is contained; in the map updating, the matching algorithm is added, instead of determining whether it is the same object only by the distance of the semantic distance; in the local map, when the number of objects exceeds a certain threshold, the matching algorithm based on the semantic distance can make the map updating more accurate.

[0142] In one embodiment, as shown in Figure 8 , a semantic map updating step is provided, which is applied to a terminal in Figure 1 for example, including the following steps:

[0143] Step 802, extracting the target object point cloud from the crowd-sourced map corresponding to the current collection area.

[0144] Step 804, constructing the first point cloud object according to the target object point cloud and extracting the corresponding first semantic information.

[0145] Step 806, determining whether there is a semantic map in the current collection area, if yes, executing step 808, otherwise, executing step 814.

[0146] Wherein, the collection area refers to the collection area where the current collection terminal is located.

[0147] Step 808, when there is a semantic map in the newly collected collection area, obtaining all the second point cloud objects and the corresponding second semantic information in the semantic map.

[0148] Step 810, matching the crowd-sourced map and the semantic map according to the first semantic information and the second semantic information, to obtain the matching result.

[0149] Specifically, according to the first semantic information and the second semantic information, the center position coordinate distance difference value, the point cloud object direction difference, the calibration frame size difference and the appearance feature difference between the first point cloud object and each second point cloud object are determined; the semantic distance is determined according to the center position coordinate distance difference value, the point cloud object direction difference value, the calibration frame size difference value and the appearance feature difference value; the number n of the first point cloud objects in the crowd-sourced map and the number m of the second point cloud objects in the semantic map are obtained, to obtain the correlation matrix of n*m;

[0150] The semantic distance is taken as an element of the association matrix; when there is a first element in the association matrix greater than or equal to a predetermined threshold, it is determined that the first point cloud object and the second point cloud object corresponding to the first element are not a point cloud object matching pair; when there is at least one second element in the association matrix less than the predetermined threshold, it is determined that there is at least one matching second point cloud object for the first point cloud object corresponding to the second element; the association matrix is segmented based on the second element to obtain a plurality of subgraphs; bipartite graph matching is performed on each subgraph to determine the matching cost value of the first point cloud object and each second point cloud object in the subgraph; and the second point cloud object corresponding to the matching cost value with the smallest value is determined as the matching point cloud object of the first point cloud object. In step 812, the semantic map is updated according to the matching result.

[0151] Specifically, when the matching result is that the first point cloud object is a newly added point cloud object, the first point cloud object is added to the semantic map; when the matching result is that the second point cloud object is a disappeared point cloud object, the second point cloud object is deleted from the semantic map; and when the matching result is that the first point cloud object has a point cloud object matching pair, fusion processing is performed on the first point cloud object and the matching point cloud object to obtain a fused point cloud object, and the fusion semantic information of the fused point cloud is determined; and the semantic map is updated according to the fusion semantic information.

[0152] In other words, when there is a matching point cloud object matching the first point cloud object in the semantic map, ICP, NDT or other algorithms are used to fuse the object point clouds corresponding to the first point cloud object and the matching point cloud object to obtain a fused point cloud, and denoising and clustering processing is performed on the fused point cloud. The semantic information is recalculated and extracted from the down-sampled fused point cloud. The recalculated and extracted semantic information of the fused point cloud can be obtained by weighted mean processing of the semantic information of the first point cloud object and the second point cloud object matching the first point cloud object. For example, the point cloud center, point cloud convex hull, point cloud minimum bounding box OBB, point cloud PCA coordinate system direction, point cloud main direction and point cloud histogram in the semantic information of the two are subjected to mean processing, and the specific manner can be realized by existing methods, which will not be described here.

[0153] In step 814, a semantic map is created according to the first semantic information.

[0154] In the above semantic map updating steps, the target first point cloud object is extracted from the newly collected crowd-sourced map, and the first semantic information of the first point cloud object is extracted, that is, under the premise of retaining useful information, the overall size of the map is greatly reduced, and more map information is obtained while the data size is compressed during the extraction of the semantic information. The matching result of the crowd-sourced map and the semantic map is determined based on the semantic distance determined based on the semantic information; and according to different matching results, the semantic map is updated by adding new objects, deleting objects and averaging matching pairs of objects, which provides a guarantee for map matching update, and further improves the map update accuracy.

[0155] In another embodiment, as shown in Figure 9 , a semantic map updating method is provided, which is described by taking a terminal in Figure 1 as an example, including the following steps:

[0156] Step 902, extracting a target object point cloud from a crowdsourcing map of a current collection area.

[0157] Step 904, constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information.

[0158] Step 906, judging whether there is a semantic map in the collection area, if yes, executing step 908, otherwise, executing step 922.

[0159] Among them, the collection area refers to the collection area where the terminal is located.

[0160] Step 908, when there is a semantic map in the newly collected collection area, obtaining all second point cloud objects and corresponding second semantic information in the semantic map.

[0161] Step 910, matching the crowdsourcing map and the semantic map according to the first semantic information and the second semantic information, and obtaining a matching result.

[0162] Step 912, when the matching result is that the first point cloud object is a new point cloud object, adding the first point cloud object to the semantic map.

[0163] Step 914, when the matching result is that the second point cloud object is a disappearing point cloud object, deleting the second point cloud object from the semantic map.

[0164] Step 916, when the matching result is that the first point cloud object has a point cloud object matching pair, obtaining a matching point cloud object matched with the first point cloud object from the second point cloud object.

[0165] Step 918, performing fusion processing on the first point cloud object and the matching point cloud object to obtain a fusion point cloud object and determine fusion semantic information of the fusion point cloud.

[0166] Step 920, updating the semantic map according to the fusion semantic information.

[0167] Step 922, when there is no semantic map in the collection area, creating a semantic map according to the first semantic information.

[0168] The following is the application scenario of the semantic map updating method, as shown in Figure 10As shown, it includes three parts of original data processing, crowd-sourcing mapping, positioning and loop detection, wherein the original data processing refers to perception on 3D LiDAR point cloud data acquired according to Light Detection And Ranging (LiDAR) and images (such as 2D camera images) collected by an image collection device (such as a camera), to obtain a newly collected crowd-sourcing map, and to extract target object point cloud from the newly collected crowd-sourcing map; such as identifying 2D camera images through a perception deep learning model, determining 2D traffic light calibration frame and 2D lane line calibration frame, projecting 3D LiDAR point cloud of the laser radar into the 2D camera image, extracting points within the traffic light calibration frame, and then restoring 3D, which are considered as points of the traffic light, to obtain point cloud corresponding to the traffic light; determining the perception identification points of the traffic sign in the point cloud through the deep learning method of the perception group, and the points marked as traffic signs can be directly taken as traffic sign point cloud; perceiving 2D lane line calibration frame in 2D camera image through a perception deep learning model, projecting 3D LiDAR point cloud of the laser radar into the 2D camera image, extracting points within the lane line calibration frame, and then restoring 3D, which are considered as points of the lane line; and further perceiving other objects with traffic signs to obtain corresponding point cloud.

[0169] According to the target object point cloud, a first point cloud object (such as object object) is constructed and corresponding first semantic information is extracted, and crowd-sourcing mapping is performed according to the constructed first point cloud object; further, it is judged whether the newly collected collection area has a map (i.e. whether there is a semantic map), if not, mapping is performed according to the first point cloud object; when there is a semantic map, the crowd-sourcing map and the existing semantic map are matched through Hungarian matching, change detection is realized, and the detection result includes added object (which can be understood as newly added point cloud object in the existing semantic map), deleted object (which can be understood as disappeared point cloud object in the existing semantic map) and average object (which can be understood as matching point cloud object pair existing in the existing semantic map and the crowd-sourcing map); the semantic map is updated according to the detection result. The specific detection method is specifically defined in the above definition of the semantic map updating method, which will not be repeated here.

[0170] Further, the positioning and loop detection are performed according to the updated semantic map, the detection capability of loop detection is improved, the cumulative error is reduced, the positioning precision and speed obstacle avoidance are improved; wherein, the positioning and loop detection method can be realized by the existing mode, and details are not described herein. For example, after obtaining the updated semantic map, the positioning and loop detection are performed on the semantic map, the objects and lane lines are matched, and the positioning result is obtained. Wherein, the loop detection, also known as the closed loop detection, refers to the ability of the device to identify a scene that has been reached, so that the map is closed. In the above semantic map updating method, the target first point cloud object is extracted from the newly collected crowd-sourced map, and the first semantic information of the first point cloud object is extracted, that is, under the premise of retaining useful information, the overall size of the map is greatly reduced, and more map information is obtained while the semantic information is extracted and the data size is compressed; the matching result of the crowd-sourced map and the semantic map is determined based on the semantic distance determined based on the semantic information; and the semantic map is updated by adding new objects, deleting objects and averaging matching objects according to different matching results, which provides a guarantee for map matching and updating, and further improves the map updating precision.

[0171] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0172] Based on the same inventive concept, the embodiments of the present application also provide a semantic map updating device for implementing the above-mentioned semantic map updating method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, and therefore the specific limitations in one or more semantic map updating device embodiments provided below can refer to the limitations of the semantic map updating method described above, and details are not described herein.

[0173] In one embodiment, as shown in Figure 11 A semantic map updating device is provided, comprising: an extraction module 1102, a construction module 1104, a first acquisition module 1106, a matching module 1108 and an updating module 1111, wherein:

[0174] The extraction module 1102 is configured to extract a target object point cloud from the crowdsourcing map of the current collection area.

[0175] The construction module 1104 is configured to construct a first point cloud object according to the target object point cloud, and the extraction module is further configured to extract first semantic information of the first point cloud object.

[0176] The first acquisition module 1106 is configured to acquire all second point cloud objects and corresponding second semantic information in a semantic map when the current collection area has a corresponding semantic map.

[0177] The matching module 1108 is configured to match the crowdsourcing map and the semantic map according to the first semantic information and the second semantic information, to obtain a matching result.

[0178] The update module 1110 is configured to update the semantic map according to the matching result.

[0179] The above semantic map updating apparatus extracts a target first point cloud object from a newly collected crowdsourcing map, and extracts first semantic information of the first point cloud object, so that the overall size of the map is greatly reduced on the premise of retaining useful information, and more map information is obtained while the semantic information is extracted and the data size is compressed; when the collection area has a semantic map, the crowdsourcing map and the semantic map are matched according to the second semantic information of all second point cloud objects in the semantic map and the first semantic information, that is, the newly collected data is matched and updated with the existing semantic map, which provides a guarantee for map matching and updating, and thus improves the map updating accuracy.

[0180] Optionally, in an embodiment, the semantic map updating apparatus further includes a determination module configured to determine a semantic distance between the first point cloud object and each second point cloud object according to the first semantic information and the second semantic information.

[0181] Optionally, in an embodiment, the matching module 1108 is further configured to match the crowdsourcing map and the semantic map based on the semantic distance, to obtain a matching result.

[0182] Optionally, in an embodiment, the determination module is further configured to determine a center position coordinate distance difference value, a point cloud object direction difference, a bounding box size difference, and an appearance feature difference between the first point cloud object and each second point cloud object according to the first semantic information and the second semantic information.

[0183] The semantic distance is determined according to the center position coordinate distance difference value, the point cloud object direction difference value, the bounding box size difference value, and the appearance feature difference value.

[0184] Optionally, in an embodiment, the determining module is further configured to determine a center position coordinate distance difference value according to the point cloud center coordinate of the first point cloud object and the point cloud center coordinate of the second point cloud object.

[0185] determine a point cloud object direction difference value according to the point cloud main direction of the first point cloud object and the point cloud main direction of the second point cloud object.

[0186] determine a calibration frame size difference value according to the calibration frame size of the first point cloud object and the calibration frame size of the second point cloud object.

[0187] determine an appearance feature difference value according to the shape element of the point cloud histogram of the first point cloud object and the shape element of the point cloud histogram of the second point cloud object.

[0188] The semantic distance between each first point cloud object and the second point cloud object in the semantic map is obtained by weighting the center position coordinate distance difference score, the point cloud object direction difference score, the calibration frame size difference score, and the appearance feature difference score.

[0189] Optionally, in an embodiment, the matching module 1108 comprises a first obtaining module, wherein:

[0190] The first obtaining module is configured to obtain the number n of the first point cloud objects in the crowd-sourced map and the number m of the second point cloud objects in the semantic map, and obtain an association matrix of n*m.

[0191] The matching module 1108 is configured to take the semantic distance as an element of the association matrix, and obtain a matching result of the first point cloud object in the crowd-sourced map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix.

[0192] The matching module 1108 is further configured to, when there is a first element in the association matrix that is greater than or equal to the predetermined threshold, determine that the first point cloud object and the second point cloud object corresponding to the first element are not a point cloud object matching pair.

[0193] The matching module 1108 comprises a segmentation module, which is configured to, when there is at least one second element in the association matrix that is less than the predetermined threshold, determine that the first point cloud object corresponding to the second element has at least one matched second point cloud object, and segment the association matrix based on the second element to obtain a plurality of sub-maps.

[0194] The matching module 1108 is further configured to perform bipartite graph matching on each sub-map to determine the second point cloud object matched with the first point cloud object.

[0195] It is further configured to perform bipartite graph matching on each sub-map to determine a matching value of the first point cloud object and each second point cloud object in the sub-map.

[0196] The second point cloud object corresponding to the smallest matching cost is determined as the matching point cloud object of the first point cloud object. Optionally, in one embodiment, the update module 1110 is further configured to add the first point cloud object to the semantic map when the matching result indicates that the first point cloud object is a newly added point cloud object.

[0197] Optionally, in one embodiment, the update module 1110 is further configured to delete the second point cloud object from the semantic map when the matching result is that the second point cloud object is a vanishing point cloud object.

[0198] Optionally, in one embodiment, the update module 1110 further includes a fusion module, wherein:

[0199] The fusion module is used to fuse the first point cloud object and the matched point cloud object to obtain the fused point cloud object, and to determine the fusion semantic information of the fused point cloud.

[0200] The update module 1110 is also used to update the semantic map based on the fused semantic information.

[0201] A module is created to generate a semantic map based on the first semantic information when a semantic map does not exist in the collection area.

[0202] Each module in the aforementioned semantic map update device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0203] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 12 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a semantic map update method. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0204] Those skilled in the art can understand that, Figure 12 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0205] In one embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0206] In one embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program implements the steps in the above method embodiments when executed by a processor.

[0207] In one embodiment, a computer program product is provided, including a computer program, and the computer program implements the steps in the above method embodiments when executed by a processor.

[0208] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0209] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0210] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0211] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A semantic map updating method characterized by, The method comprises: extracting a target object point cloud from a crowdsourcing map corresponding to a current collection area; constructing a first point cloud object according to the target object point cloud and extracting corresponding first semantic information; when there is a corresponding semantic map in the current collection area, obtaining all second point cloud objects in the semantic map and corresponding second semantic information; determining a center position coordinate distance difference value, a point cloud object direction difference value, a calibration box size difference value and an appearance feature difference value between the first point cloud object and each second point cloud object according to the first semantic information and the second semantic information; determining a semantic distance according to the center position coordinate distance difference value, the point cloud object direction difference value, the calibration box size difference value and the appearance feature difference value; matching the crowdsourcing map and the semantic map based on the semantic distance to obtain a matching result, wherein the matching result is used to represent whether the first point cloud object has a matched second point cloud object in the semantic map; updating the semantic map according to the matching result.

2. The method of claim 1, wherein, The first semantic information and the second semantic information comprise a point cloud center coordinate, a point cloud calibration box size, a point cloud main direction and a point cloud histogram, and the determination of the center position coordinate distance difference value, the point cloud object direction difference value, the calibration box size difference value and the appearance feature difference value between the first point cloud object and each second point cloud object according to the first semantic information and the second semantic information comprises: determining the center position coordinate distance difference value according to the point cloud center coordinate of the first point cloud object and the point cloud center coordinate of the second point cloud object; determining the point cloud object direction difference value according to the point cloud main direction of the first point cloud object and the point cloud main direction of the second point cloud object; determining the calibration box size difference value according to the calibration box size of the first point cloud object and the calibration box size of the second point cloud object; determining the appearance feature difference value according to the shape elements of the point cloud histogram of the first point cloud object and the shape elements of the point cloud histogram of the second point cloud object.

3. The method of claim 1, wherein, The determination of the semantic distance according to the center position coordinate distance difference value, the point cloud object direction difference value, the calibration box size difference value and the appearance feature difference value comprises: performing weighted processing on the center position coordinate distance difference value, the point cloud object direction difference value, the calibration box size difference value and the appearance feature difference value to obtain a semantic distance between each first point cloud object and a second point cloud object in the semantic map.

4. The method of claim 1, wherein, The matching of the crowdsourcing map and the semantic map based on the semantic distance to obtain a matching result comprises: obtaining the number n of first point cloud objects in the crowdsourcing map and the number m of second point cloud objects in the semantic map to obtain an association matrix of n*m; taking the semantic distance as an element of the association matrix; obtaining a matching result of the first point cloud object in the crowdsourcing map and the second point cloud object in the semantic map according to a predetermined threshold and the association matrix.

5. The method of claim 4, wherein, The matching result of the first point cloud object in the crowdsourcing map and the second point cloud object in the semantic map is obtained according to a predetermined threshold and the correlation matrix, and the method comprises the following steps: When there is a first element greater than or equal to the predetermined threshold in the correlation matrix, it is determined that the first point cloud object and the second point cloud object corresponding to the first element are not a point cloud object matching pair.

6. The method of claim 4, wherein, The matching result of the first point cloud object in the crowdsourcing map and the second point cloud object in the semantic map is obtained according to a predetermined threshold and the correlation matrix, and the method comprises the following steps: When there is at least one second element less than the predetermined threshold in the correlation matrix, it is determined that the first point cloud object corresponding to the second element has at least one matching second point cloud object; The correlation matrix is segmented based on the second element to obtain a plurality of sub-maps; A bipartite graph matching is performed on each of the sub-maps to determine the second point cloud object matching the first point cloud object.

7. The method of claim 6, wherein, The bipartite graph matching is performed on each of the sub-maps to determine the second point cloud object matching the first point cloud object, and the method comprises the following steps: A bipartite graph matching is performed on each of the sub-maps to determine the matching cost value of the first point cloud object and each of the second point cloud objects in the sub-map; The second point cloud object corresponding to the matching cost value with the smallest value is determined as the matching point cloud object of the first point cloud object.

8. The method of claim 5, wherein, The semantic map is updated according to the matching result, and the method comprises the following steps: When the matching result is that the first point cloud object is a new point cloud object, the first point cloud object is added to the semantic map.

9. The method of claim 5, wherein, The semantic map is updated according to the matching result, and the method comprises the following steps: When the matching result is that the second point cloud object is a disappearing point cloud object, the second point cloud object is deleted from the semantic map.

10. The method of claim 7, wherein, The semantic map is updated according to the matching result, and the method comprises the following steps: The first point cloud object and the matching point cloud object are fused to obtain a fused point cloud object, and the fused semantic information of the fused point cloud is determined; The semantic map is updated according to the fused semantic information.

11. The method of claim 1, wherein, The method further comprises the following steps: When there is no semantic map in the current collection area, a semantic map is created according to the first semantic information.

12. A semantic map updating apparatus characterized by comprising: The device comprises the following modules: An extraction module is configured to extract a target object point cloud from a crowdsourcing map corresponding to a current collection area; A construction module is configured to construct a first point cloud object according to the target object point cloud; the extraction module is further configured to extract first semantic information of the first point cloud object; A first acquisition module is configured to acquire all second point cloud objects in a semantic map corresponding to the current collection area and corresponding second semantic information when the current collection area has the corresponding semantic map; A second acquisition module is configured to acquire a correlation matrix of the first point cloud object and the second point cloud object in the semantic map corresponding to the current collection area. The matching module is configured to determine a center position coordinate distance difference value, a point cloud object direction difference value, a bounding box size difference value, and an appearance feature difference value between the first point cloud object and each of the second point cloud objects according to the first semantic information and the second semantic information; determine a semantic distance according to the center position coordinate distance difference value, the point cloud object direction difference value, the bounding box size difference value, and the appearance feature difference value; and match the crowd-sourced map and the semantic map based on the semantic distance to obtain a matching result, wherein the matching result is used to represent whether the first point cloud object has a matched second point cloud object in the semantic map. The updating module is configured to update the semantic map according to the matching result.

13. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 11.

14. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 11.

Images