Semantic map optimization method and device

Abstract

The invention relates to the technical field of map construction, and discloses a semantic map optimization method and device, and the method comprises the steps: determining a to-be-optimized local map sequence; the local map sequence comprises a plurality of local maps; each local map comprises a pose index of the local map and a plurality of point cloud data under a local coordinate system of the local map; based on the pose index of each local map, performing local optimization operation on the local map sequence to obtain a local optimization result; based on the local optimization result, loopback detection operation is executed, and a loopback detection result is obtained; based on the local optimization result and the loopback detection result, executing global optimization operation to obtain a global optimization result; the global optimization result comprises a global optimization index of each local map. Therefore, by implementing the method, the optimization efficiency and the optimization accuracy of the semantic map can be improved, so that the accuracy of the semantic map is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of map construction, and in particular to a method and device for optimizing a semantic map. Background Art

[0002] At present, based on SLAM (Simultaneous Localization and Mapping) technology, data collected by sensors such as lidar and cameras can be processed to obtain point cloud data to construct semantic maps. Among them, for large-scale scenes, it is necessary to fuse multiple point cloud maps of the scene to obtain a complete semantic map of the scene.

[0003] In the prior art, if a vehicle equipped with sensors is used to collect environmental data in a scene (for example, a parking lot scene in smart parking technology), the vehicle's odometer must be used for analysis when building a map. However, in actual applications, as the distance traveled by the vehicle increases when collecting data, the accuracy of the vehicle's odometer will gradually decrease, causing the accuracy of the semantic map built by the vehicle's odometer to decrease, resulting in a reduction in the accuracy of the subsequent positioning cruise using the constructed semantic map.

[0004] It can be seen that it is particularly important to propose a technical solution that can improve the optimization efficiency and optimization accuracy of semantic maps, thereby improving the accuracy of semantic maps, so as to improve the accuracy of subsequent positioning and cruising using semantic maps. Summary of the invention

[0005] The present invention provides a semantic map optimization method and device, which can improve the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, which is beneficial to improving the accuracy of subsequent positioning and cruising using the semantic map.

[0006] In order to solve the above technical problems, the first aspect of the present invention discloses a method for optimizing a semantic map, the method comprising:

[0007] Determine a local map sequence to be optimized; the local map sequence includes multiple local maps; each local map includes a pose index of the local map and multiple point cloud data in a local coordinate system of the local map;

[0008] Based on the posture index of each of the local maps, a local optimization operation is performed on the local map sequence to obtain a local optimization result; the local optimization result includes the local optimization index of each of the local maps;

[0009] Based on the local optimization result, a loop detection operation is performed to obtain a loop detection result;

[0010] Based on the local optimization results and the loop detection results, a global optimization operation is performed to obtain a global optimization result; the global optimization result includes a global optimization index of each of the local maps.

[0011] As an optional implementation, in the first aspect of the present invention, performing a local optimization operation on the local map sequence based on the pose index of each of the local maps to obtain a local optimization result includes:

[0012] For each of the local maps, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculate the pose change amount between the local map and the adjacent map;

[0013] For each of the local maps, according to the amount of position change, each of the point cloud data of the local map is converted to the local coordinate system of the adjacent map to obtain first converted point cloud data corresponding to each of the point cloud data included in the local map;

[0014] For each of the local maps, determining a first overlapping area between the local map and the adjacent map;

[0015] For each of the local maps, the posture index of the local map is optimized according to the posture change amount, each of the first converted point cloud data and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map.

[0016] As an optional embodiment, in the first aspect of the present invention, for each of the local maps, the point cloud data corresponding to the first overlapping area includes a plurality of first overlapping point cloud data contained in the local map regarding the first overlapping area and a plurality of second overlapping point cloud data contained in the adjacent map regarding the first overlapping area.

[0017] As an optional implementation, in the first aspect of the present invention, for each of the local maps, optimizing the pose index of the local map according to the pose change amount and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map includes:

[0018] Determine first converted point cloud data corresponding to each of the first overlapping point cloud data as first overlapping converted data;

[0019] Matching corresponding second overlapping point cloud data for each of the first overlapping conversion data to obtain a plurality of first matching point cloud data pairs corresponding to the first overlapping area;

[0020] Based on a predetermined error formula, determining a posture optimization amount corresponding to the local map according to the posture index of the local map and all the first matching point cloud data pairs;

[0021] According to the posture change amount and the posture optimization amount, the posture index of the local map is optimized to obtain a local optimization index of the local map.

[0022] As an optional implementation, in the first aspect of the present invention, the loop detection result includes at least one loop local map pair;

[0023] The performing of a loop detection operation based on the local optimization result to obtain a loop detection result includes:

[0024] For each of the local maps, based on a predetermined distance calculation formula, the index distance between the local optimization index of the local map and the local optimization index of each previously ranked map is calculated; each of the previously ranked maps is a local map ranked earlier than the local map in the local map sequence;

[0025] For each of the local maps, when it is detected that the minimum index distance among all the index distances is less than or equal to a preset distance threshold, the sorted previous map corresponding to the minimum index distance and the local map are determined as a looped local map pair.

[0026] As an optional implementation, in the first aspect of the present invention, each of the loop closure local map pairs includes a first loop closure map and a second loop closure map; the loop closure detection result also includes a plurality of second matching point cloud data pairs of each of the loop closure local map pairs;

[0027] The performing of a loop detection operation based on the local optimization result to obtain a loop detection result further includes:

[0028] For each of the loop closure local map pairs, convert each of the point cloud data included in the first loop closure map into the local coordinate system of the second loop closure map to obtain second converted point cloud data corresponding to each of the point cloud data included in the first loop closure map;

[0029] For each of the loop closure local map pairs, corresponding point cloud data is matched for each of the second converted point cloud data from the second loop closure map to obtain a plurality of second matching point cloud data pairs of the loop closure local map pair.

[0030] As an optional implementation manner, in the first aspect of the present invention, performing a global optimization operation based on the local optimization result and the loop detection result to obtain a global optimization result includes:

[0031] Based on a predetermined global optimization formula, the global optimization index of each local map is calculated according to the pose indexes of all the local maps, all the local optimization indexes in the local optimization results, and all the second matching point cloud data pairs in the loop detection results.

[0032] A second aspect of the present invention discloses a semantic map optimization device, the device comprising:

[0033] A determination module, used to determine a local map sequence to be optimized; the local map sequence includes multiple local maps; each of the local maps includes a pose index of the local map and multiple point cloud data in a local coordinate system of the local map;

[0034] A local optimization module, configured to perform a local optimization operation on the local map sequence based on the posture index of each local map to obtain a local optimization result; the local optimization result includes the local optimization index of each local map;

[0035] A loop detection module, used to perform a loop detection operation based on the local optimization result to obtain a loop detection result;

[0036] A global optimization module is used to perform a global optimization operation based on the local optimization results and the loop detection results to obtain a global optimization result; the global optimization result includes a global optimization index of each of the local maps.

[0037] As an optional implementation, in the second aspect of the present invention, the local optimization module performs a local optimization operation on the local map sequence based on the pose index of each of the local maps, and the specific manner of obtaining the local optimization result includes:

[0038] For each of the local maps, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculate the pose change amount between the local map and the adjacent map;

[0039] For each of the local maps, according to the amount of position change, each of the point cloud data of the local map is converted to the local coordinate system of the adjacent map to obtain first converted point cloud data corresponding to each of the point cloud data included in the local map;

[0040] For each of the local maps, determining a first overlapping area between the local map and the adjacent map;

[0041] For each of the local maps, the posture index of the local map is optimized according to the posture change amount, each of the first converted point cloud data and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map.

[0042] As an optional embodiment, in the second aspect of the present invention, for each of the local maps, the point cloud data corresponding to the first overlapping area includes a plurality of first overlapping point cloud data contained in the local map regarding the first overlapping area and a plurality of second overlapping point cloud data contained in the adjacent map regarding the first overlapping area.

[0043] As an optional implementation, in the second aspect of the present invention, for each of the local maps, the local optimization module optimizes the pose index of the local map according to the pose change amount and the point cloud data corresponding to the first overlapping area, and the specific method of obtaining the local optimization index of the local map includes:

[0044] Determine first converted point cloud data corresponding to each of the first overlapping point cloud data as first overlapping converted data;

[0045] Matching corresponding second overlapping point cloud data for each of the first overlapping conversion data to obtain a plurality of first matching point cloud data pairs corresponding to the first overlapping area;

[0046] Based on a predetermined error formula, determining a posture optimization amount corresponding to the local map according to the posture index of the local map and all the first matching point cloud data pairs;

[0047] According to the posture change amount and the posture optimization amount, the posture index of the local map is optimized to obtain a local optimization index of the local map.

[0048] As an optional implementation, in the second aspect of the present invention, the loop detection result includes at least one loop local map pair;

[0049] The loop detection module performs a loop detection operation based on the local optimization result, and a specific method for obtaining the loop detection result includes:

[0050] For each of the local maps, based on a predetermined distance calculation formula, the index distance between the local optimization index of the local map and the local optimization index of each previously ranked map is calculated; each of the previously ranked maps is a local map ranked earlier than the local map in the local map sequence;

[0051] For each of the local maps, when it is detected that the minimum index distance among all the index distances is less than or equal to a preset distance threshold, the sorted previous map corresponding to the minimum index distance and the local map are determined as a looped local map pair.

[0052] As an optional implementation, in the second aspect of the present invention, each of the loop closure local map pairs includes a first loop closure map and a second loop closure map; the loop closure detection result also includes a plurality of second matching point cloud data pairs of each of the loop closure local map pairs;

[0053] The loop detection module performs a loop detection operation based on the local optimization result, and a specific method of obtaining the loop detection result also includes:

[0054] For each of the loop closure local map pairs, convert each of the point cloud data included in the first loop closure map into the local coordinate system of the second loop closure map to obtain second converted point cloud data corresponding to each of the point cloud data included in the first loop closure map;

[0055] For each of the loop closure local map pairs, corresponding point cloud data is matched for each of the second converted point cloud data from the second loop closure map to obtain a plurality of second matching point cloud data pairs of the loop closure local map pair.

[0056] As an optional implementation, in the second aspect of the present invention, the global optimization module performs a global optimization operation based on the local optimization result and the loop detection result, and a specific manner of obtaining the global optimization result includes:

[0057] Based on a predetermined global optimization formula, the global optimization index of each local map is calculated according to the pose indexes of all the local maps, all the local optimization indexes in the local optimization results, and all the second matching point cloud data pairs in the loop detection results.

[0058] A third aspect of the present invention discloses another semantic map optimization device, the device comprising:

[0059] A memory storing executable program code;

[0060] a processor coupled to the memory;

[0061] The processor calls the executable program code stored in the memory to execute the semantic map optimization method disclosed in the first aspect of the present invention.

[0062] The fourth aspect of the present invention discloses a computer storage medium, wherein the computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the semantic map optimization method disclosed in the first aspect of the present invention.

[0063] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0064] In an embodiment of the present invention, a local map sequence to be optimized is determined; the local map sequence includes multiple local maps; each local map includes a posture index of the local map and multiple point cloud data in the local coordinate system of the local map; based on the posture index of each local map, a local optimization operation is performed on the local map sequence to obtain a local optimization result; the local optimization result includes the local optimization index of each local map; based on the local optimization result, a loop detection operation is performed to obtain a loop detection result; based on the local optimization result and the loop detection result, a global optimization operation is performed to obtain a global optimization result; the global optimization result includes the global optimization index of each local map. It can be seen that the implementation of the present invention can first locally optimize the local map sequence based on the posture index of each local map, then perform loop detection based on the local optimization results, and then combine the local optimization results and the loop detection results to globally optimize all local maps to obtain the global optimization index of each local map. Compared with the operation of optimizing the map after the loop is detected in the prior art, the method can locally optimize the map without being constrained by the loop detection, and can also further globally optimize the map by combining the local optimization results and the loop detection results, thereby improving the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, which is beneficial to improving the accuracy of subsequent positioning and cruising using the semantic map. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0066] Figure 1 It is a flowchart of a semantic map optimization method disclosed in an embodiment of the present invention;

[0067] Figure 2 It is a flowchart of another method for optimizing a semantic map disclosed in an embodiment of the present invention;

[0068] Figure 3 It is a schematic diagram of the principle of a semantic map optimization method disclosed in an embodiment of the present invention;

[0069] Figure 4 It is a schematic diagram of the principle of another semantic map optimization method disclosed in an embodiment of the present invention;

[0070] Figure 5 It is a schematic diagram of the principle of another semantic map optimization method disclosed in an embodiment of the present invention;

[0071] Figure 6 It is a flowchart of another method for optimizing a semantic map disclosed in an embodiment of the present invention;

[0072] Figure 7 It is a structural schematic diagram of a semantic map optimization device disclosed in an embodiment of the present invention;

[0073] Figure 8 It is a structural schematic diagram of another semantic map optimization device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0076] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0077] The present invention discloses a semantic map optimization method and device, which can first locally optimize the local map sequence based on the posture index of each local map, then perform loop detection based on the local optimization results, and then combine the local optimization results and the loop detection results to globally optimize all local maps to obtain the global optimization index of each local map. Compared with the operation of optimizing the map after detecting the loop in the prior art, this method can locally optimize the map without being constrained by the loop detection, and can also comprehensively optimize the map with the local optimization results and the loop detection results, thereby improving the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, which is conducive to improving the accuracy of the subsequent positioning cruise using the semantic map. The following are detailed descriptions.

[0078] Embodiment 1

[0079] See also Figure 1 , Figure 1 : is a flow chart of a semantic map optimization method disclosed in an embodiment of the present invention. Figure 1 The semantic map optimization method described above can be applied to a semantic map optimization device, which can include an optimization device, an optimization terminal, an optimization system, and a server, wherein the server can be a local server or a cloud server, which is not limited in the embodiments of the present invention. Figure 1 As shown, the optimization method of the semantic map may include the following operations:

[0080] 101. Determine a local map sequence to be optimized.

[0081] In an embodiment of the present invention, a local map sequence may include multiple local maps; optionally, the local map sequence may be sorted based on the creation time of each local map, which is not limited by the embodiment of the present invention; wherein each local map may include a posture index of the local map and multiple point cloud data in the local coordinate system of the local map, and the point cloud data may include semantic information; optionally, when the local map sequence is collected by a vehicle-mounted sensor (for example, a vehicle body camera), the posture index of each local map may be the vehicle body posture when the local map is created, which is not limited by the embodiment of the present invention.

[0082] In an embodiment of the present invention, the local coordinate system of each local map is established based on the posture index of the local map; optionally, the local coordinate system of each local map can be a right-handed coordinate system. For example, the center of the vehicle body posture when the local map is created can be the origin, the forward direction of the vehicle body can be the positive direction of the Y axis, the right side of the vehicle body can be the positive direction of the X axis, and the upward direction can be the positive direction of the Z axis. This embodiment of the present invention is not limited.

[0083] In the embodiment of the present invention, for example, the pose index of each local map can be pose k =(R k ,t k ), specifically:

[0084]

[0085] R k =R yaw ·R roll ·R ptich

[0086] t k =[x k ;y k ; z k ]

[0087] Among them, R k It represents the rotation matrix, which is composed of the yaw angle, the pitch angle, and the roll angle. The unit is radian. The above yaw angle, pitch angle, and roll angle are the angles of the vehicle itself relative to the global coordinate system corresponding to the current scanning scene. k Represents the position of the vehicle itself in the global coordinate system, represented by x k ,y k ,z k The present invention does not limit the composition.

[0088] 102. Based on the pose index of each local map, a local optimization operation is performed on the local map sequence to obtain a local optimization result.

[0089] In the embodiment of the present invention, the local optimization result may include a local optimization index of each local map.

[0090] 103. Based on the local optimization result, a loop detection operation is performed to obtain a loop detection result.

[0091] In the embodiment of the present invention, the loop detection operation is used to detect whether there are at least two local maps that form a loop in the local map sequence.

[0092] In the embodiment of the present invention, optionally, the loop detection operation may be performed after each local map is locally optimized, or the loop detection operation may be performed after all local maps are locally optimized, which is not limited in the embodiment of the present invention.

[0093] 104. Based on the local optimization results and the loop detection results, a global optimization operation is performed to obtain a global optimization result.

[0094] In the embodiment of the present invention, the global optimization result may include a global optimization index of each local map.

[0095] In the embodiment of the present invention, optionally, after step 104, the method may further include the following operations:

[0096] Based on the global optimization results, all local maps are spliced ​​to obtain the global map.

[0097] It can be seen that the method described in the embodiment of the present invention can first locally optimize the local map sequence based on the posture index of each local map, then perform loop detection based on the local optimization results, and then combine the local optimization results and the loop detection results to globally optimize all local maps to obtain the global optimization index of each local map. Compared with the operation of optimizing the map after the loop is detected in the prior art, this method can locally optimize the map without being constrained by loop detection, and can also further globally optimize the map by combining the local optimization results and the loop detection results, thereby improving the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, which is beneficial to improving the accuracy of subsequent positioning and cruising using the semantic map.

[0098] In an optional embodiment, when using a vehicle-mounted sensor to collect real-time point cloud data, determining a local map sequence to be optimized may include the following operations:

[0099] After each local map is created, it is checked whether the real-time point cloud data exceeds the area range of the current local map; when it is detected that the real-time point cloud data exceeds the area range of the current local map, a new local map is created and the above-mentioned operation of checking whether the real-time point cloud data exceeds the area range of the current local map is re-executed.

[0100] For example, when each local map is composed of 3×3 point cloud data, the area range of the local map is the area range corresponding to the above point cloud data; assuming that the current local map is a local Figure 1 , and the real-time point cloud data exceeds the local Figure 1 If the area is large, new local land Figure 2 At this time, the real-time point cloud data is in the overlapping area of ​​the two local maps. Figure 1 , locally Figure 2 and real-time point cloud data (i.e. Figure 3 The current frame point cloud shown in Figure 3 As shown, Figure 3 It is a schematic diagram of the principle of a semantic map optimization method disclosed in an embodiment of the present invention.

[0101] It can be seen that this optional embodiment can create a new local map when the point cloud data collected in real time exceeds the area range of the current local map, thereby improving the efficiency of creating the local map.

[0102] In an optional embodiment, based on the pose index of each local map, a local optimization operation is performed on the local map sequence to obtain a local optimization result, which may include the following operations:

[0103] For each local map, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculate the pose change between the local map and the adjacent map;

[0104] For each local map, each point cloud data of the local map is converted to the local coordinate system of the adjacent map according to the amount of position change, so as to obtain first converted point cloud data corresponding to each point cloud data included in the local map;

[0105] For each local map, determining a first overlapping area between the local map and an adjacent map;

[0106] For each local map, the pose index of the local map is optimized according to the pose change amount, each first converted point cloud data and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map.

[0107] Among them, optionally, the adjacent map corresponding to each local map can be a local map that is ranked adjacent to or adjacent to the local map in the local map sequence. Exemplarily, the adjacent map can be a previous local map created before the local map, which is not limited in the embodiment of the present invention.

[0108] Optionally, for each local map, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculating the pose change amount between the local map and the adjacent map may include the following operations:

[0109] Assume that the local map is MASK k+1 , the adjacent map corresponding to this local map is MASK k , then the specific calculation method of the position change between the local map and the adjacent map can be as follows:

[0110] ΔR k,k+1 =R k T ·R k+1

[0111] Δt k,k+1 =R k T ·(t k+1 -t k )

[0112] Among them, ΔR k,k+1 and Δt k,k+1is the position change between the local map and the adjacent map, (R k+1 ,t k+1 ) is the pose index of the local map, (R k ,t k ) is the pose index of the adjacent map.

[0113] Optionally, for each local map, each point cloud data of the local map is converted to the local coordinate system of the adjacent map according to the amount of position change, so as to obtain first converted point cloud data corresponding to each point cloud data included in the local map, which may include the following operations:

[0114] Use local map MASK k+1 For example, the adjacent map corresponding to the local map is MASK k , then the specific conversion method of each point cloud data conversion coordinate system can be as follows:

[0115]

[0116] Among them, P k+1 j , j = 1…N is the local map MASK k+1 Point cloud data, P k ' j P k+1 j Switch to adjacent map MASK k The first transformed point cloud data obtained after the local coordinate system, ΔR k,k+1 and Δt k,k+1 is the above-mentioned posture change.

[0117] It can be seen that this optional embodiment can convert the point cloud data of the local map to the local coordinate system of the adjacent map based on the calculated pose change between the pose index of the local map and the pose index of the adjacent map, and then optimize the pose index of the local map according to the pose change, each first converted point cloud data and the point cloud data corresponding to the first overlapping area between the local map and the adjacent map to obtain a local optimized index, thereby improving the local optimization accuracy of the map pose index.

[0118] In this optional embodiment, optionally, for each local map, the point cloud data corresponding to the first overlapping area may include multiple first overlapping point cloud data contained in the local map regarding the first overlapping area and multiple second overlapping point cloud data contained in the adjacent map regarding the first overlapping area.

[0119] In this optional embodiment, optionally, for each local map, optimizing the pose index of the local map according to the pose change amount and the point cloud data corresponding to the first overlapping area to obtain the local optimized index of the local map may include the following operations:

[0120] Determine the first converted point cloud data corresponding to each first overlapping point cloud data as the first overlapping converted data;

[0121] Matching corresponding second overlapping point cloud data for each first overlapping conversion data to obtain a plurality of first matching point cloud data pairs corresponding to the first overlapping area;

[0122] Based on a predetermined error formula, according to the pose index of the local map and all first matching point cloud data pairs, determining a pose optimization amount corresponding to the local map;

[0123] According to the pose change amount and the pose optimization amount, the pose index of the local map is optimized to obtain the local optimization index of the local map.

[0124] The posture optimization amount corresponding to each local map may be an optimization amount of a posture change amount with respect to the local map.

[0125] Among them, exemplary ones are Figure 4 As shown, Figure 4 is a schematic diagram of another semantic map optimization method disclosed in an embodiment of the present invention. Taking local map k as an example, the first overlapping area between local map k and local map k-1 is as follows: Figure 4 The first overlapping area between the local map k and the local map k+1 is as shown in the “k overlapping area”. Figure 4 As shown in the “k+1 overlapping area”, when the local map k+1 needs to be locally optimized, all the first overlapping conversion data of the local map k+1 regarding the “k+1 overlapping area” need to be paired with the second overlapping point cloud data contained in the local map k regarding the “k+1 overlapping area”.

[0126] Optionally, based on a preset number of iterations, the pose index of the local map can be iteratively optimized for a preset number of iterations to obtain the pose optimization amount corresponding to the local map, wherein the iterative optimization operation can include the above-mentioned "matching the corresponding second overlapping point cloud data for each first overlapping conversion data to obtain multiple first matching point cloud data pairs corresponding to the first overlapping area; based on a predetermined error formula, determining the pose optimization amount corresponding to the local map according to the pose index of the local map and all first matching point cloud data pairs", which is not limited in the embodiments of the present invention. This can improve the accuracy of determining the pose optimization amount, thereby facilitating improving the optimization accuracy of the pose index of each local map.

[0127] Further optionally, the first matching point cloud data pair may be in the form of:

[0128] (P k j P k ' j ),j=1…N

[0129] Among them, P k ' j is the first overlapping conversion data, P k j The second overlapping point cloud data corresponds to the first overlapping conversion data.

[0130] Optionally, the error formula can be as follows:

[0131]

[0132] Among them, (δR k,k+1 ,δt k,k+1 ) is the change in posture ΔR k,k+1 , Δt k,k+1 The amount of optimization.

[0133] Optionally, according to the pose change amount and the pose optimization amount, the pose index of the local map is optimized to obtain the local optimization index of the local map, which can be expressed as follows:

[0134]

[0135] in, and MASK is the local map k+1 Locally optimized index.

[0136] It can be seen that this optional embodiment can also match the first overlapping conversion data of the converted point cloud data in the first overlapping area with the corresponding second overlapping point cloud data to obtain a first matching point cloud data pair, and then calculate the posture optimization amount corresponding to the local map based on the error formula, and then optimize the posture index of the local map according to the posture change amount and the posture optimization amount to obtain the local optimization index, which can further improve the local optimization accuracy of the map posture index, thereby improving the accuracy of the local optimization index.

[0137] Embodiment 2

[0138] See also Figure 2 , Figure 2 : is a flow chart of a semantic map optimization method disclosed in an embodiment of the present invention. Figure 2The semantic map optimization method described above can be applied to a semantic map optimization device, which can include an optimization device, an optimization terminal, an optimization system, and a server, wherein the server can be a local server or a cloud server, which is not limited in the embodiments of the present invention. Figure 2 As shown, the optimization method of the semantic map may include the following operations:

[0139] 201. Determine a local map sequence to be optimized.

[0140] In an embodiment of the present invention, the local map sequence includes multiple local maps; each local map includes a posture index of the local map and multiple point cloud data in the local coordinate system of the local map.

[0141] 202. Based on the pose index of each local map, a local optimization operation is performed on the local map sequence to obtain a local optimization result.

[0142] In the embodiment of the present invention, the local optimization result includes a local optimization index of each local map.

[0143] 203. For each local map, based on a predetermined distance calculation formula, calculate the index distance between the local optimization index of the local map and the local optimization index of each previously sorted map.

[0144] In the embodiment of the present invention, each of the first-order maps is a local map that is ordered earlier than the local map in the local map sequence.

[0145] In the embodiment of the present invention, optionally, a local map MASK k2 For example, the distance calculation formula can be as follows:

[0146]

[0147] Among them, dist i MASK is the local map k2 The index distance between the index of and the index of each previously sorted map, i is the serial number of the previously sorted map, and the specific meanings of x, y, and z can be referred to the detailed description of the posture index in Example 1, which will not be repeated in this embodiment of the present invention.

[0148] 204. For each local map, when it is detected that the minimum index distance among all index distances is less than or equal to a preset distance threshold, a previously sorted map corresponding to the minimum index distance and the local map are determined as a looped local map pair.

[0149] In the embodiment of the present invention, optionally, a local map MASK k2 For example, the minimum index distance among all index distances can be determined as follows:

[0150]

[0151] When dist min ≤dist thd When the dist min The corresponding sorting map MASK k1 With the local map MASK k2 Determined as a loop local map pair, where dist thd is the preset distance threshold.

[0152] In an embodiment of the present invention, the loop detection result includes at least one loop local map pair; wherein a second overlapping area exists between two local maps included in each loop local map pair and a loop is formed.

[0153] For example, it is assumed that the first loop map included in a loop local map pair is MASK k2 , The second loop map is MASK k1 ,like Figure 5 As shown, Figure 5 is a schematic diagram of the principle of another semantic map optimization method disclosed in an embodiment of the present invention. The second overlapping area between the loop local map pair is as follows: Figure 5 The dotted area is shown in .

[0154] 205. Based on the local optimization results and the loop detection results, a global optimization operation is performed to obtain a global optimization result.

[0155] In the embodiment of the present invention, the global optimization result includes the global optimization index of each local map.

[0156] In the embodiment of the present invention, for other detailed descriptions of step 201-step 202 and step 205, please refer to the detailed description of step 101-step 102 and step 104 in the first embodiment, and the embodiment of the present invention will not be repeated here.

[0157] It can be seen that the method described in the embodiment of the present invention can firstly optimize the local map sequence locally based on the pose index of each local map, then perform loop detection based on the local optimization result, and then combine the local optimization result and the loop detection result to perform global optimization on all local maps to obtain the global optimization index of each local map. Compared with the operation of optimizing the map after detecting the loop in the prior art, the method can optimize the map locally without being constrained by the loop detection, and can also combine the local optimization result and the loop detection result to further optimize the map globally, thereby improving the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, and facilitating the subsequent use of the semantic map for positioning and cruising. In addition, the index distance between the local map and each ranked map can be calculated. If the minimum index distance is detected to be less than or equal to the preset distance threshold, the local map corresponding to the local minimum index is determined as a loop local map pair, which can improve the efficiency and accuracy of the loop detection, thereby facilitating the accuracy of the global optimization.

[0158] In an optional embodiment, each loop closure local map pair may include a first loop closure map and a second loop closure map; the loop closure detection result may also include a plurality of second matching point cloud data pairs of each loop closure local map pair;

[0159] Wherein, based on the local optimization result, a loop detection operation is performed to obtain a loop detection result, which may also include the following operations:

[0160] For each loop closure local map pair, each point cloud data included in the first loop closure map is converted to the local coordinate system of the second loop closure map to obtain second converted point cloud data corresponding to each point cloud data included in the first loop closure map;

[0161] For each loop closure local map pair, corresponding point cloud data is matched for each second converted point cloud data from the second loop closure map to obtain multiple second matching point cloud data pairs of the loop closure local map pair.

[0162] For example, assuming that the first loop map of a loop local map pair is MASK k2 , The second loop map is MASK k1 , then the second matching point cloud data pair of the loop local map pair can be:

[0163] (P k1 j P k1 ' j ),j=1…N

[0164] in, is the second converted point cloud data corresponding to each point cloud data included in the first loop map, P k1 j For each second transformed point cloud data, the point cloud data is matched in the second loop map.

[0165] It can be seen that this optional embodiment can convert the point cloud data of one of the maps in the loop local map pair into the local coordinate system of the other map, and match the point cloud data in the two maps to obtain a second matching point cloud data pair, which can improve the matching accuracy of the point cloud data of the loop map, thereby improving the accuracy of the loop detection results.

[0166] In this optional embodiment, optionally, performing a global optimization operation based on the local optimization result and the loop detection result to obtain a global optimization result may include the following operations:

[0167] Based on a predetermined global optimization formula, the global optimization index of each local map is calculated according to the pose indexes of all local maps, all local optimization indexes in the local optimization results, and all second matching point cloud data pairs in the loop detection results.

[0168] Optionally, when there are K local maps that have been locally optimized, the original pose index of each local map is (R k ,t k ), the local optimization index of each local map is When there are M loop local map pairs after the loop detection operation, each loop local map pair includes {(R i ,t i ),(R j ,t j )} m ,m=1…M;i,j=1…K, the number of point cloud data of each loop local map pair is L m Then the point cloud data in each loop local map in each loop local map pair is in, represents the point cloud position of the point cloud data in the local coordinate system of the corresponding local map, and the parameter to be optimized is λ * = {R k ,t k}, k = 1…K, that is, the global optimization index of each local map, then the global optimization formula can be as follows:

[0169]

[0170] For example, a local map sequence includes local maps k1, k2, and k3. Assume that k1 and k2 are adjacent local maps, k1 and k3 are a pair of loop local maps obtained in the loop detection operation, and assume that the first matching point cloud data pair between k1 and k2 is k i (i=1,…N), N represents the number of matching point clouds between k1 and k2; assuming that the second matching point cloud data pair between k1 and k3 is k j (j=1,…M), M represents the number of matching point clouds between k1 and k3. At this time, the global optimization formula is as follows:

[0171]

[0172] It can be seen that this optional embodiment can not only use the point cloud data of the loop overlapping area as a constraint, but also add the overlapping point clouds between continuous local maps as constraints, optimize the indexes of all local maps under the loop path, so as to perform global optimization on the local maps. It can improve the global optimization accuracy of the map pose index on the basis of improving the loop optimization accuracy, thereby improving the accuracy of the global optimization index, and then improving the accuracy of the semantic map.

[0173] In the embodiment of the present invention, exemplarily, when the process of the semantic map optimization method is as follows Figure 6 When shown, Figure 6 A flowchart of another method for optimizing a semantic map disclosed in an embodiment of the present invention is shown below. The method may include the following operations:

[0174] After a new local map is created, if the number of local maps is greater than 1, the pose index of the new local map can be locally optimized. After all local maps of the scene are created, loop detection is performed on all local maps that have been locally optimized. Then, the pose indexes of all local maps are globally optimized by combining the local optimization results and the loop detection results. Optionally, all the globally optimized local maps can be spliced ​​together to obtain a global map, which is not limited in the embodiments of the present invention.

[0175] Embodiment 3

[0176] See also Figure 7 , Figure 7 1 is a schematic diagram of a structure of a semantic map optimization device disclosed in an embodiment of the present invention. Figure 7 The described semantic map optimization device may include one of an optimization device, an optimization terminal, an optimization system, and a server, wherein the server may be a local server or a cloud server, which is not limited in the embodiment of the present invention. Figure 7 As shown, the semantic map optimization device may include:

[0177] The determination module 301 is used to determine a local map sequence to be optimized; the local map sequence includes multiple local maps; each local map includes a pose index of the local map and multiple point cloud data in a local coordinate system of the local map;

[0178] A local optimization module 302 is used to perform a local optimization operation on the local map sequence based on the posture index of each local map to obtain a local optimization result; the local optimization result includes the local optimization index of each local map;

[0179] A loop detection module 303 is used to perform a loop detection operation based on the local optimization result to obtain a loop detection result;

[0180] The global optimization module 304 is used to perform a global optimization operation based on the local optimization results and the loop detection results to obtain a global optimization result; the global optimization result includes a global optimization index of each local map.

[0181] It can be seen that the device described in the embodiment of the present invention can first locally optimize the local map sequence based on the posture index of each local map, and then perform loop detection based on the local optimization results, and then combine the local optimization results and the loop detection results to globally optimize all local maps to obtain the global optimization index of each local map. Compared with the operation of optimizing the map after the loop is detected in the prior art, the method can locally optimize the map without being constrained by the loop detection, and can also further globally optimize the map by combining the local optimization results and the loop detection results, thereby improving the optimization efficiency and optimization accuracy of the semantic map, thereby improving the accuracy of the semantic map, which is beneficial to improving the accuracy of subsequent positioning and cruising using the semantic map.

[0182] In an optional embodiment, the local optimization module 302 performs a local optimization operation on the local map sequence based on the posture index of each local map, and a specific method of obtaining the local optimization result may include:

[0183] For each local map, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculate the pose change between the local map and the adjacent map;

[0184] For each local map, each point cloud data of the local map is converted to the local coordinate system of the adjacent map according to the amount of position change, so as to obtain first converted point cloud data corresponding to each point cloud data included in the local map;

[0185] For each local map, determining a first overlapping area between the local map and an adjacent map;

[0186] For each local map, the pose index of the local map is optimized according to the pose change amount, each first converted point cloud data and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map.

[0187] It can be seen that the device described in implementing this optional embodiment can convert the point cloud data of the local map to the local coordinate system of the adjacent map based on the calculated pose change between the pose index of the local map and the pose index of the adjacent map, and then optimize the pose index of the local map according to the pose change, each first converted point cloud data and the point cloud data corresponding to the first overlapping area between the local map and the adjacent map to obtain a local optimized index, thereby improving the local optimization accuracy of the map pose index.

[0188] In this optional embodiment, optionally, for each local map, the point cloud data corresponding to the first overlapping area may include multiple first overlapping point cloud data contained in the local map regarding the first overlapping area and multiple second overlapping point cloud data contained in the adjacent map regarding the first overlapping area.

[0189] It can be seen that implementing the device described in this optional embodiment can also improve the accuracy of the point cloud data corresponding to the first overlapping area.

[0190] In this optional embodiment, optionally, for each local map, the local optimization module 302 optimizes the pose index of the local map according to the pose change amount and the point cloud data corresponding to the first overlapping area, and the specific method of obtaining the local optimization index of the local map may include:

[0191] Determine the first converted point cloud data corresponding to each first overlapping point cloud data as the first overlapping converted data;

[0192] Matching corresponding second overlapping point cloud data for each first overlapping conversion data to obtain a plurality of first matching point cloud data pairs corresponding to the first overlapping area;

[0193] Based on a predetermined error formula, according to the pose index of the local map and all first matching point cloud data pairs, determining a pose optimization amount corresponding to the local map;

[0194] According to the pose change amount and the pose optimization amount, the pose index of the local map is optimized to obtain the local optimization index of the local map.

[0195] It can be seen that the device described in the implementation of this optional embodiment can also match the first overlapping conversion data of the conversion point cloud data in the first overlapping area with the corresponding second overlapping point cloud data to obtain a first matching point cloud data pair, and then calculate the posture optimization amount corresponding to the local map based on the error formula, and then optimize the posture index of the local map according to the posture change amount and the posture optimization amount to obtain the local optimization index, which can further improve the local optimization accuracy of the map posture index, thereby improving the accuracy of the local optimization index.

[0196] In another optional embodiment, the loop detection result includes at least one loop local map pair;

[0197] The loop detection module 303 performs a loop detection operation based on the local optimization result, and a specific method of obtaining the loop detection result may include:

[0198] For each local map, based on a predetermined distance calculation formula, the index distance between the local optimization index of the local map and the local optimization index of each previously ranked map is calculated; each previously ranked map is a local map ranked earlier than the local map in the local map sequence;

[0199] For each local map, when it is detected that the minimum index distance among all index distances is less than or equal to a preset distance threshold, the sorted previous map corresponding to the minimum index distance and the local map are determined as a looped local map pair.

[0200] It can be seen that the device described in the implementation of this optional embodiment is capable of calculating the index distance between the local map and each previously sorted map. If it is detected that the minimum index distance is less than or equal to the preset distance threshold, the local map and the local map corresponding to the minimum index are determined as a loop local map pair, which can improve the efficiency and accuracy of loop detection, thereby facilitating improving the accuracy of global optimization.

[0201] In this optional embodiment, optionally, each loop closure local map pair includes a first loop closure map and a second loop closure map; the loop closure detection result also includes a plurality of second matching point cloud data pairs of each loop closure local map pair;

[0202] The loop detection module 303 performs a loop detection operation based on the local optimization result, and a specific method of obtaining the loop detection result may also include:

[0203] For each loop closure local map pair, each point cloud data included in the first loop closure map is converted to the local coordinate system of the second loop closure map to obtain second converted point cloud data corresponding to each point cloud data included in the first loop closure map;

[0204] For each loop closure local map pair, corresponding point cloud data is matched for each second converted point cloud data from the second loop closure map to obtain multiple second matching point cloud data pairs of the loop closure local map pair.

[0205] It can be seen that the device described in implementing this optional embodiment can also convert the point cloud data of one of the maps in the loop local map pair into the local coordinate system of another map, and match the point cloud data in the two maps to obtain a second matching point cloud data pair, which can improve the matching accuracy of the point cloud data of the loop map, thereby improving the accuracy of the loop detection results.

[0206] In this optional embodiment, optionally, the global optimization module 304 performs a global optimization operation based on the local optimization result and the loop detection result. The specific method of obtaining the global optimization result may include:

[0207] Based on a predetermined global optimization formula, the global optimization index of each local map is calculated according to the pose indexes of all local maps, all local optimization indexes in the local optimization results, and all second matching point cloud data pairs in the loop detection results.

[0208] It can be seen that the device described in the implementation of this optional embodiment can not only use the point cloud data of the loop overlapping area as a constraint, but also add the overlapping point clouds between continuous local maps as constraints, optimize the indexes of all local maps under the loop path, so as to perform global optimization on the local maps. It can improve the global optimization accuracy of the map pose index on the basis of improving the loop optimization accuracy, thereby improving the accuracy of the global optimization index, and then improving the accuracy of the semantic map.

[0209] Embodiment 4

[0210] See also Figure 8 , Figure 8 FIG. 1 is a schematic diagram of the structure of another semantic map optimization device disclosed in an embodiment of the present invention. Figure 8 As shown, the semantic map optimization device may include:

[0211] A memory 401 storing executable program codes;

[0212] a processor 402 coupled to the memory 401;

[0213] The processor 402 calls the executable program code stored in the memory 401 to execute the steps of the semantic map optimization method described in the first embodiment of the present invention or the second embodiment of the present invention.

[0214] Embodiment 5

[0215] An embodiment of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the steps in the semantic map optimization method described in Embodiment 1 or Embodiment 2 of the present invention.

[0216] Embodiment 6

[0217] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the semantic map optimization method described in Example 1 or Example 2.

[0218] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0219] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0220] Finally, it should be noted that the semantic map optimization method and device disclosed in the embodiments of the present invention disclose only the preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing a semantic map, characterized in that: The method comprises: Determine a local map sequence to be optimized; the local map sequence includes multiple local maps; each local map includes a pose index of the local map and multiple point cloud data in a local coordinate system of the local map; Based on the posture index of each of the local maps, a local optimization operation is performed on the local map sequence to obtain a local optimization result; the local optimization result includes the local optimization index of each of the local maps; Based on the local optimization result, a loop detection operation is performed to obtain a loop detection result; Based on the local optimization results and the loop detection results, a global optimization operation is performed to obtain a global optimization result; the global optimization result includes a global optimization index of each of the local maps.

2. The method for optimizing a semantic map according to claim 1, characterized in that: The performing of a local optimization operation on the local map sequence based on the pose index of each local map to obtain a local optimization result includes: For each of the local maps, according to the pose index of the local map and the pose index of the adjacent map corresponding to the local map, calculate the pose change amount between the local map and the adjacent map; For each of the local maps, according to the amount of position change, each of the point cloud data of the local map is converted to the local coordinate system of the adjacent map to obtain first converted point cloud data corresponding to each of the point cloud data included in the local map; For each of the local maps, determining a first overlapping area between the local map and the adjacent map; For each of the local maps, the posture index of the local map is optimized according to the posture change amount, each of the first converted point cloud data and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map.

3. The method for optimizing a semantic map according to claim 2, characterized in that: For each of the local maps, the point cloud data corresponding to the first overlapping area includes a plurality of first overlapping point cloud data contained in the local map regarding the first overlapping area and a plurality of second overlapping point cloud data contained in the adjacent map regarding the first overlapping area.

4. The method for optimizing a semantic map according to claim 3, characterized in that: For each of the local maps, optimizing the posture index of the local map according to the posture change amount and the point cloud data corresponding to the first overlapping area to obtain a local optimized index of the local map includes: Determine first converted point cloud data corresponding to each of the first overlapping point cloud data as first overlapping converted data; Matching corresponding second overlapping point cloud data for each of the first overlapping conversion data to obtain a plurality of first matching point cloud data pairs corresponding to the first overlapping area; Based on a predetermined error formula, determining a posture optimization amount corresponding to the local map according to the posture index of the local map and all the first matching point cloud data pairs; According to the posture change amount and the posture optimization amount, the posture index of the local map is optimized to obtain a local optimization index of the local map.

5. The method for optimizing a semantic map according to any one of claims 1 to 4, characterized in that: The loop detection result includes at least one loop local map pair; The performing of a loop detection operation based on the local optimization result to obtain a loop detection result includes: For each of the local maps, based on a predetermined distance calculation formula, the index distance between the local optimization index of the local map and the local optimization index of each previously ranked map is calculated; each of the previously ranked maps is a local map ranked earlier than the local map in the local map sequence; For each of the local maps, when it is detected that the minimum index distance among all the index distances is less than or equal to a preset distance threshold, the sorted previous map corresponding to the minimum index distance and the local map are determined as a looped local map pair.

6. The method for optimizing a semantic map according to claim 5, characterized in that: Each of the loop closure local map pairs includes a first loop closure map and a second loop closure map; the loop closure detection result also includes a plurality of second matching point cloud data pairs of each of the loop closure local map pairs; The performing of a loop detection operation based on the local optimization result to obtain a loop detection result further includes: For each of the loop closure local map pairs, convert each of the point cloud data included in the first loop closure map into the local coordinate system of the second loop closure map to obtain second converted point cloud data corresponding to each of the point cloud data included in the first loop closure map; For each of the loop closure local map pairs, corresponding point cloud data is matched for each of the second converted point cloud data from the second loop closure map to obtain a plurality of second matching point cloud data pairs of the loop closure local map pair.

7. The method for optimizing a semantic map according to claim 6, characterized in that: The performing a global optimization operation based on the local optimization result and the loop detection result to obtain a global optimization result includes: Based on a predetermined global optimization formula, the global optimization index of each local map is calculated according to the pose indexes of all the local maps, all the local optimization indexes in the local optimization results, and all the second matching point cloud data pairs in the loop detection results.

8. A semantic map optimization device, characterized in that: The device comprises: A determination module, used to determine a local map sequence to be optimized; the local map sequence includes multiple local maps; each of the local maps includes a pose index of the local map and multiple point cloud data in a local coordinate system of the local map; A local optimization module, configured to perform a local optimization operation on the local map sequence based on the posture index of each local map to obtain a local optimization result; the local optimization result includes the local optimization index of each local map; A loop detection module, used to perform a loop detection operation based on the local optimization result to obtain a loop detection result; A global optimization module is used to perform a global optimization operation based on the local optimization results and the loop detection results to obtain a global optimization result; the global optimization result includes a global optimization index of each of the local maps.

9. A semantic map optimization device, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the semantic map optimization method as described in any one of claims 1-7.

10. A computer storage medium, characterized in that: The computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the semantic map optimization method as described in any one of claims 1-7.

Images