Model processing and high-precision map quality checking method and device and storage medium

Abstract

The present disclosure provides a model processing and high-definition map quality verification method and device and a storage medium, relating to the fields of artificial intelligence such as automatic driving, machine learning and computer vision, wherein the model processing method can comprise: constructing a Bayesian network model; determining the values of each variable in the Bayesian network model according to the obtained training samples; and determining the network parameters of the Bayesian network model according to the training samples and the values of each variable, wherein the network parameters are used to determine the quality verification result of the high-definition map to be verified, and the quality verification result is determined according to the input data and the network parameters determined based on the point cloud data corresponding to the high-definition map. The scheme disclosed in the present disclosure can save labor and time costs and improve the accuracy of the verification result.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of artificial intelligence, in particular to a model processing and high-definition map quality checking method and device and storage medium in the fields of automatic driving, machine learning and computer vision. BACKGROUND

[0002] High-definition maps play an important role in the automatic driving scene, and the quality of high-definition maps directly affects the safety of automatic driving, etc., therefore, the quality of high-definition maps needs to be checked. SUMMARY

[0003] The present disclosure provides a model processing and high-definition map quality checking method, device and storage medium.

[0004] A model processing method comprises:

[0005] constructing a Bayesian network model;

[0006] determining the values of each variable in the Bayesian network model according to the obtained training samples;

[0007] determining network parameters of the Bayesian network model according to the training samples and the values of each variable, the network parameters being used to determine a quality checking result of a high-definition map to be checked, the quality checking result being determined according to input data determined based on a point cloud data pair corresponding to the high-definition map and the network parameters, the point cloud data being point cloud data used to generate the high-definition map.

[0008] A high-definition map quality checking method comprises:

[0009] for a high-definition map to be checked, obtaining at least one group of point cloud data pairs, the point cloud data being point cloud data used to generate the high-definition map;

[0010] for each group of point cloud data pairs, respectively performing the following processing: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality checking result of the point cloud data pair according to the input data and network parameters of a Bayesian network model determined according to training samples;

[0011] determining a quality checking result of the high-definition map according to the quality checking results of each point cloud data pair.

[0012] A model processing device comprises a model construction module, a variable determination module and a parameter determination module.

[0013] The model construction module is configured to construct a Bayesian network model.

[0014] The variable determination module is configured to determine values of variables in the Bayesian network model according to the obtained training samples.

[0015] The parameter determination module is configured to determine network parameters of the Bayesian network model according to the training samples and the values of the variables, the network parameters being used to determine a quality checking result of the high-definition map to be checked, the quality checking result being determined according to input data determined based on a point cloud data pair corresponding to the high-definition map and the network parameters.

[0016] A high-definition map quality checking device, comprising a data obtaining module, a first determination module and a second determination module.

[0017] The data obtaining module is configured to obtain at least one group of point cloud data pairs for a high-definition map to be checked, the point cloud data being used to generate the high-definition map.

[0018] The first determination module is configured to, for each group of point cloud data pairs, perform the following processing: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality checking result of the point cloud data pair according to the input data and network parameters of a Bayesian network model determined according to training samples.

[0019] The second determination module is configured to determine a quality checking result of the high-definition map according to the quality checking results of the point cloud data pairs.

[0020] An electronic device, comprising:

[0021] at least one processor; and

[0022] a memory communicatively connected to the at least one processor; wherein

[0023] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0024] A non-transitory computer-readable storage medium storing computer instructions, the computer instructions being used to cause a computer to perform the method as described above.

[0025] A computer program product, comprising computer programs / instructions, the computer programs / instructions being executed by a processor to implement the method as described above.

[0026] It should be understood that the matters described in this section are not intended to identify key or essential features of the embodiments of the present disclosure, nor are they used to limit the scope of the present disclosure. Other features of the present disclosure will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0027] The accompanying drawings are used to better understand the present scheme, and do not constitute a limitation on the present disclosure. Among them:

[0028] Figure 1 Flow chart of the model processing method embodiment described in the present disclosure;

[0029] Figure 2 Structural diagram of the Bayesian network model described in the present disclosure;

[0030] Figure 3 Flow chart of the high-definition map quality verification method embodiment described in the present disclosure;

[0031] Figure 4 Structural diagram of the model processing device embodiment 400 described in the present disclosure;

[0032] Figure 5 Structural diagram of the high-definition map quality verification device embodiment 500 described in the present disclosure;

[0033] Figure 6 A schematic block diagram of an electronic device 600 that can be used to implement embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0034] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to help understanding, and should be considered only as exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0035] In addition, it should be understood that the term "and / or" herein is only a description of the association relationship between the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after it.

[0036] In the scheme described in the present disclosure, a high-definition map quality verification method based on a Bayesian network model is proposed. First, the related model processing method is described below.

[0037] Figure 1A flowchart of an embodiment of the model processing method according to the present disclosure is shown in FIG. 1. As shown in FIG. 1, the embodiment includes the following specific implementation. Figure 1

[0038] In step 101, a Bayesian network model is constructed.

[0039] In step 102, values of variables in the Bayesian network model are determined according to the obtained training samples.

[0040] In step 103, network parameters of the Bayesian network model are determined according to the training samples and the values of the variables, the network parameters being used to determine a quality checking result of the high-definition map to be checked, the quality checking result being determined according to input data determined based on point cloud data corresponding to the high-definition map and the network parameters, the point cloud data being point cloud data used to generate the high-definition map.

[0041] In the conventional manner, the quality of the high-definition map is usually checked manually, which requires a large amount of manpower and time cost, and has poor accuracy.

[0042] By using the scheme according to the above method embodiment, the Bayesian network model can be constructed, and the network parameters of the Bayesian network model can be determined according to the training samples, and then the quality checking of the high-definition map can be automatically realized by means of the network parameters, thereby saving the manpower and time cost, improving the processing efficiency, and improving the accuracy of the checking result, etc.

[0043] In one embodiment of the present disclosure, each training sample can include M kinds of predetermined data, M being a positive integer greater than one, and correspondingly, the variables included in the Bayesian network model can include variables corresponding to the M kinds of predetermined data respectively.

[0044] The number of training samples is not limited, and in general, the more the number, the better the subsequent processing effect.

[0045] In one embodiment of the present disclosure, each training sample can correspond to a group of point cloud data pairs, and the M kinds of predetermined data can include data difference values corresponding to two frames of point cloud data in the point cloud data pair, relative poses between the two frames of point cloud data, first intermediate results, second intermediate results, third intermediate results, height ghosting, and level ghosting, wherein the relative poses can include x-axis difference values, y-axis difference values, z-axis difference values, pitch difference values, roll difference values, yaw difference values, height ghosting (HG, Height Ghosting), and level ghosting (LG, Level Ghosting) are used to reflect the quality of the point cloud data pair, and the first intermediate results, the second intermediate results, and the third intermediate results are all intermediate results determined according to the data difference values and the relative poses.​

[0046] For example, for a certain area in Beijing, actual point cloud data can be collected on the roads in the area, such as using a laser radar on a collection vehicle to collect multiple frames of point cloud data, and then a high-definition map of the area can be generated based on the collected multiple frames of point cloud data. Similarly, for other areas, corresponding high-definition maps can also be generated.

[0047] Based on each high-definition map, the training samples can be obtained. For example, at least one set of point cloud data pairs can be determined from the point cloud data corresponding to the high-definition map, each point cloud data pair can include two frames of point cloud data, and the two frames of point cloud data in each set of point cloud data pairs can be two frames of point cloud data that meet a predetermined relationship. As one possible implementation, they can be two adjacent key frames, in which a frame at an intermediate position within a certain range can be selected as a key frame, and the two key frames corresponding to the adjacent range are adjacent key frames. For each set of point cloud data pairs, a training sample can be obtained, such as by actual calculation and in combination with manual annotation, etc.

[0048] Each training sample can include: a data difference value corresponding to the two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, a third intermediate result, a height ghosting, and a horizontal ghosting, wherein the relative pose can include: an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value.

[0049] For example, for each key frame, the point cloud data of other frames within the same range can be fused into the coordinate system of the key frame according to the relative pose to the key frame, to form a point cloud denser local map corresponding to each key frame. For the two frames of point cloud data in the point cloud data pair, the point cloud data in the corresponding local map can be processed, such as by a convolutional neural network for dimension reduction processing, mapping to a one-dimensional variable, and then the absolute value of the difference between the two one-dimensional variables can be obtained as the required data difference value.

[0050] In addition, the relative pose can include 6 degrees of freedom, which belong to the difference in translation and the difference in rotation direction, respectively. The difference in translation includes the x-axis difference value, the y-axis difference value, and the z-axis difference value, i.e., the difference in the x-axis direction, the difference in the y-axis direction, and the difference in the z-axis direction. The difference in rotation direction includes the pitch angle difference value, the roll angle difference value, and the yaw angle difference value.

[0051] Based on the data difference and the relative pose, a first intermediate result, a second intermediate result and a third intermediate result can be further determined. The first intermediate result can be referred to as ground thickness, the second intermediate result can be referred to as road edge linearity, and the third intermediate result can be referred to as wall surface flatness. How to determine the intermediate results according to the data difference and the relative pose is not limited. For example, the intermediate results can be determined according to a predetermined formula or through data fitting.

[0052] The height ghosting and the horizontal ghosting are used to represent the quality of the point cloud data, and since the high-precision map is generated according to the point cloud data, the quality of the high-precision map can be determined by analyzing the quality of the point cloud data.

[0053] As can be seen, the predetermined data includes various data related to the quality verification of the high-precision map, thereby laying a good foundation for subsequent processing.

[0054] As described above, the variables included in the Bayesian network model can include variables corresponding to the M predetermined data. In an embodiment of the disclosure, the Bayesian network model can include first layer variables, second layer variables and third layer variables. The number of the first layer variables can be 7, the number of the second layer variables can be 3, and the number of the third layer variables can be 2, that is, the value of M can be 12.

[0055] Specifically, in an embodiment of the disclosure, the first layer variables can include a first variable corresponding to the data difference, a second variable corresponding to the x-axis difference, a third variable corresponding to the y-axis difference, a fourth variable corresponding to the z-axis difference, a fifth variable corresponding to the pitch angle difference, a sixth variable corresponding to the roll angle difference, and a seventh variable corresponding to the yaw angle difference, the second layer variables can include an eighth variable corresponding to the first intermediate result, a ninth variable corresponding to the second intermediate result, and a tenth variable corresponding to the third intermediate result, and the third layer variables can include an eleventh variable corresponding to the height ghosting and a twelfth variable corresponding to the horizontal ghosting.

[0056] In addition, in an embodiment of the disclosure, the parent nodes of the eighth variable can include the first variable, the fourth variable, the fifth variable, the sixth variable and the seventh variable, each variable is a node, that is, the variable can also be referred to as a node, the parent nodes of the ninth variable can include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable and the seventh variable, the parent nodes of the tenth variable can include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable and the seventh variable, the parent nodes of the eleventh variable can include the eighth variable, and the parent nodes of the twelfth variable can include the ninth variable and the tenth variable.

[0057] That is, a Bayesian network model can be constructed based on each variable and the causal relationship between each variable, etc., so as to realize high-precision map quality verification based on the constructed Bayesian network model, and improve the access efficiency of the high-precision map.

[0058] In combination with the above description, Figure 2 is a structural schematic diagram of the Bayesian network model of the present disclosure. As shown in the figure, Figure 2 the first variable to the twelfth variable can also be represented in the following manner: PCD, X, Y, Z, PITCH, ROLL, YAW, GT, CS, TF, HG, and LG, wherein HG is mainly related to GT, and LG is mainly related to CS and TF.

[0059] According to the training samples, the values of each variable in the Bayesian network model can be determined. In an embodiment of the present disclosure, for any variable, the corresponding discrete values of the variable can be determined, and each value (probability value) in the discrete values corresponds to a different interval range.

[0060] In theory, the values of each variable are continuous values. In order to simplify the calculation and facilitate subsequent calculation, the values of each variable can be discretized.

[0061] In an embodiment of the present disclosure, for any variable, the manner of determining the corresponding discrete values thereof can include: for each interval range corresponding to the variable, the following processing is performed respectively: the number of training samples in which the predetermined data corresponding to the variable is located in the interval range is counted, the ratio of the counted number of training samples to the total number of training samples is obtained, and the ratio is taken as the value corresponding to the interval range.

[0062] For any variable, how to divide the interval range is not limited. For example, taking the second variable as an example, it is assumed that the following interval ranges are divided: [0, 1), [1, 2), [2, 3), [3, 4), and [4, +∞). In addition, it is assumed that a total of 100 (this number is only for illustration, and in practice it can be much larger) training samples are obtained, of which 10 training samples have x-axis difference values located in the interval range [0, 1), and then the value corresponding to the interval range [0, 1) is 1 / 10 (i.e. 10 / 100, which can also be denoted as 10%), and it is assumed that another 20 training samples have x-axis difference values located in the interval range [1, 2), and then the value corresponding to the interval range [1, 2) is 1 / 5, and the others are similar.

[0063] The above is described by taking the second variable as an example, and the processing manner of other variables is the same.

[0064] It can be seen that, by the above manner, the values corresponding to different interval ranges of different variables can be determined efficiently and accurately based on the training samples, thereby realizing the discretization processing of the values of the variables, i.e., the values of the variables required can be obtained.

[0065] Further, the network parameters of the Bayesian network model, i.e., the model parameters, can be determined according to the training samples and the values of the variables.

[0066] In an embodiment of the present disclosure, the training samples can be preprocessed, and the preprocessing can include: for any predetermined data in any training sample, determining the variable corresponding to the predetermined data respectively, and determining the interval range to which the predetermined data belongs from the interval ranges corresponding to the corresponding variables, and replacing the predetermined data with the value corresponding to the interval range to which the predetermined data belongs, and then the network parameters of the Bayesian network model can be determined according to the preprocessed training samples and the values of the variables.

[0067] That is, each predetermined data in the training samples can be mapped into a discrete value of a variable, thereby facilitating subsequent processing, for example, the x-axis difference in a certain training sample is 1.7, which is located in the interval range [1, 2), and then the value corresponding to the interval range (such as 1 / 5) can be used to replace the x-axis difference in the training sample.

[0068] In an embodiment of the present disclosure, the network parameters of the Bayesian network model can be determined by a maximum likelihood estimation (also referred to as a maximum likelihood estimation) manner according to the preprocessed training samples and the values of the variables.

[0069] For the Bayesian network model, it can be decomposed into local probability models according to conditional independence, and each local probability model can be solved respectively, i.e., for a single parameter, it can be solved by a maximum likelihood estimation manner.

[0070] For example:

[0071] For each value of GT, the values of each θ can be obtained by maximizing the likelihood function.

[0072]

[0073] When the value of GT is gt0, the above maximum likelihood function can be represented as:

[0074]

[0075] Wherein, pcd0, z0, pitch0, roll0 and yaw0 represent the values of PCD, Z, PITCH, ROLL and YAW respectively, and assuming that the values (discretized values) of GT include gt0, gt1, gt2 and gt3, the values of each θ can be obtained for each value.

[0076] When the GT takes the value of gt0, the values of each θ can be obtained by maximizing the likelihood function as follows:

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] wherein M[,,] represents a sufficient statistic corresponding to the value of the variable, and For example, M[gt0,pcd0] represents the number of training samples in the preprocessed training samples with the first intermediate result gt0 and the data difference value pcd0, and M[GT,pcd0] represents the number of training samples in the preprocessed training samples with the data difference value pcd0.

[0083] Each θ shown in (3) to (7) is the network parameter required, and other network parameters of the Bayesian network model can be obtained in a similar manner, that is, the learning of the Bayesian network parameters can be completed by the maximum likelihood estimation method, wherein for each value of each variable shown in Figure 2 Each value of each variable needs to determine the corresponding θ, the first layer variable has no parent node, which is an edge variable, and for example, PCD can directly obtain and the like.

[0084] The maximum likelihood estimation method is a mature existing algorithm, and accordingly, the maximum likelihood estimation method is used to determine the network parameters of the Bayesian network model, which can improve the accuracy of the determination result.

[0085] After the network parameters are determined, they can be applied to the actual high-definition map quality verification, that is, for the high-definition map to be verified, the determined input data and the network parameters can be determined based on the point cloud data corresponding to the high-definition map to determine the quality verification result of the high-definition map, and the point cloud data is the point cloud data used to generate the high-definition map.

[0086] Correspondingly, Figure 3 is a flowchart of the high-definition map quality verification method embodiment of the present disclosure. As Figure 3 shown, it includes the following specific implementation methods.

[0087] In step 301, at least one set of point cloud data pairs is obtained for the high-definition map to be verified, and the point cloud data is the point cloud data used to generate the high-definition map.

[0088] In step 302, for each group of point cloud data pairs, the following processing is performed respectively: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality check result of the point cloud data pair according to the input data and network parameters of a Bayesian network model determined according to a training sample.

[0089] In step 303, a quality check result of the high-definition map is determined according to the quality check results of each point cloud data pair.

[0090] By adopting the scheme described in the above method embodiment, the Bayesian network model can be constructed in advance, and the network parameters of the Bayesian network model can be determined according to the training sample. Accordingly, the quality check of the high-definition map can be automatically implemented by means of the network parameters, thereby saving the labor and time cost, improving the processing efficiency, and improving the accuracy of the check result, etc.

[0091] The number of groups of point cloud data pairs that are specifically obtained can be determined according to actual needs.

[0092] For each group of point cloud data pairs, input data can be determined according to two frames of point cloud data therein. In an embodiment of the present disclosure, the input data can include: a data difference value corresponding to the two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, and a third intermediate result, the relative pose can include: an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value, the first intermediate result, the second intermediate result, and the third intermediate result are all intermediate results determined according to the data difference value and the relative pose between the two frames of point cloud data.

[0093] After the input data is obtained, a quality check result of the point cloud data pair can be determined according to the input data and network parameters of a Bayesian network model determined according to a training sample.

[0094] Specifically, in an embodiment of the present disclosure, a first calculation expression and a second calculation expression can be respectively determined according to the Bayesian network model by means of variable elimination method, then a height ghosting of the point cloud data pair can be determined according to the input data, the network parameters, and the first calculation expression, a horizontal ghosting of the point cloud data pair can be determined according to the input data, the network parameters, and the second calculation expression, and finally the quality check result of the point cloud data pair can be determined according to the determined height ghosting and horizontal ghosting.

[0095] That is, the expression of the quality of the point cloud data pair (i.e., the quality of the high-definition map) can be obtained by means of the variable elimination method, and the variable elimination method is a mature existing algorithm. Accordingly, based on the obtained expression, the required height ghosting and horizontal ghosting can be efficiently and accurately determined.

[0096] The first calculation expression can be:

[0097] P'(HG) = P(HG|GT)P(GT|PCD)P(GT|Z)P(GT|PITCH)P(GT|ROLL)P(GT|YAW)P(PCD)P(Z)P(PITCH)P(ROLL)P(YAW); (8)

[0098] The second calculation expression can be:

[0099] P'(LG) = P(LG|CS)P(CS|PCD)P(CS|X)P(CS|Y)P(CS|Z)P(CS|PITCH)P(CS|ROLL)P(CS|YAW)P(LG|TF)P(TF|PCD)P(TF|X)P(TF|Y)P(TF|Z)P(TF|PITCH)P(TF|ROLL)P(TF|YAW)P(PCD)P(X)P(Y)P(Z)P(PITCH)P(ROLL)P(YAW); (9)

[0100] Wherein, P'(HG) represents the first calculation result, and P'(LG) represents the second calculation result. Taking the formula (8) as an example, P(GT|PCD) in the formula (8) represents a value corresponding to the first intermediate result and the difference value data in the input data, for example, assuming that the first intermediate result in the input data is mapped to a value corresponding to a range of intervals, and the value obtained after mapping is gt0, and the difference value data in the input data is mapped to a value corresponding to a range of intervals, and the value obtained after mapping is pcd0, then P(GT|PCD) is equal to the value of and other analogies.

[0101] Still taking the formula (8) as an example, for P(HG|GT) in the formula (8), since HG is not included in the input data, the values (discrete values) corresponding to different interval ranges of HG are respectively taken as the values of HG, and are substituted into the formula (8) for calculation. Correspondingly, assuming that HG corresponds to 5 different interval ranges, then 5 P'(HG) can be obtained, and similarly, multiple P'(LG) can be obtained.

[0102] Further, the P'(HG) with the largest value can be selected from the multiple P'(HG) obtained, as the height ghosting, or the value of HG corresponding to the P'(HG) with the largest value can be taken as the height ghosting. In addition, the P'(LG) with the largest value can be selected from the multiple P'(LG) obtained, as the horizontal ghosting, or the value of LG corresponding to the P'(LG) with the largest value can be taken as the horizontal ghosting.

[0103] That is, the quality judgment can be converted into the maximum possible interpretation of the problem, that is, the value with the maximum probability is considered as the most possible result.

[0104] Further, the obtained high ghosting and horizontal ghosting can be compared with the respective corresponding threshold values respectively, and if greater than the threshold value, it can be considered that there is a quality problem, such as the existence of high ghosting and / or horizontal ghosting problem, otherwise, it can be considered that there is no quality problem.

[0105] In addition, the quality checking result of the high-precision map can also be determined according to the quality checking result of each point cloud data pair. For example, if the quality checking result of each point cloud data pair is that there is no quality problem, it can be considered that the high-precision map has no quality problem, otherwise, it can be considered that the high-precision map has a quality problem, and targeted optimization can be performed for the existing quality problem.

[0106] It should be noted that for the foregoing method embodiments, in order to simply describe, they are all expressed as a series of action combinations, but those skilled in the art should know that the disclosure is not limited by the order of the described actions, because according to the disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the disclosure. In addition, the parts not described in detail in a certain embodiment can refer to the related description in other embodiments.

[0107] In summary, by using the scheme described in the method embodiment of the disclosure, the quality checking of the high-precision map can be automatically realized, thereby saving the manpower and time cost, improving the processing efficiency, and improving the accuracy of the checking result.

[0108] The above is the introduction of the method embodiment, and the scheme described in the disclosure will be further described through the device embodiment.

[0109] Figure 4 The constituent structure schematic diagram of the model processing device embodiment 400 of the disclosure is shown. As shown in Figure 4 It includes a model construction module 401, a variable determination module 402, and a parameter determination module 403.

[0110] The model construction module 401 is configured to construct a Bayesian network model.

[0111] The variable determination module 402 is configured to determine the value of each variable in the Bayesian network model according to the obtained training sample.

[0112] The parameter determination module 403 is configured to determine network parameters of the Bayesian network model according to the training samples and values of the variables, and the network parameters are used to determine a quality verification result of the high-definition map to be verified, and the quality verification result is determined according to input data determined based on point cloud data corresponding to the high-definition map and the network parameters, and the point cloud data is point cloud data used to generate the high-definition map.

[0113] By adopting the scheme of the device embodiment, the Bayesian network model can be constructed, the network parameters of the Bayesian network model can be determined according to the training samples, and then the quality verification of the high-definition map can be automatically implemented by means of the network parameters, thereby saving the labor and time costs, improving the processing efficiency, and improving the accuracy of the verification result.

[0114] In an embodiment of the present disclosure, each training sample can include M kinds of predetermined data respectively, and M is a positive integer greater than one. Correspondingly, the variables included in the Bayesian network model can include variables corresponding to the M kinds of predetermined data respectively.

[0115] In an embodiment of the present disclosure, each training sample can correspond to a group of point cloud data pairs, and the M kinds of predetermined data can include data difference values corresponding to two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, a third intermediate result, a height ghosting, and a horizontal ghosting. The relative pose can include an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value. The height ghosting and the horizontal ghosting are used to reflect the quality of the point cloud data pair. The first intermediate result, the second intermediate result, and the third intermediate result are all intermediate results determined according to the data difference values and the relative pose.

[0116] As described above, the variables included in the Bayesian network model can include variables corresponding to the M kinds of predetermined data respectively. In an embodiment of the present disclosure, the Bayesian network model can include first layer variables, second layer variables, and third layer variables. The number of the first layer variables can be 7, the data amount of the second layer variables can be 3, and the number of the third layer variables can be 2.

[0117] Specifically, in an embodiment of the present disclosure, the first layer variables can include a first variable corresponding to the data difference value, a second variable corresponding to the x-axis difference value, a third variable corresponding to the y-axis difference value, a fourth variable corresponding to the z-axis difference value, a fifth variable corresponding to the pitch angle difference value, a sixth variable corresponding to the roll angle difference value, and a seventh variable corresponding to the yaw angle difference value. The second layer variables can include an eighth variable corresponding to the first intermediate result, a ninth variable corresponding to the second intermediate result, and a tenth variable corresponding to the third intermediate result. The third layer variables can include an eleventh variable corresponding to the height ghosting and a twelfth variable corresponding to the horizontal ghosting.

[0118] In addition, in one embodiment of the present disclosure, the parent node of the eighth variable can include the first variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable, each of which is a node, i.e., the variable can also be referred to as a node, the parent node of the ninth variable can include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable, the parent node of the tenth variable can include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable, and the parent node of the eleventh variable can include the eighth variable, and the parent node of the twelfth variable can include the ninth variable and the tenth variable.

[0119] According to the training samples, the values of each variable in the Bayesian network model can be determined. In one embodiment of the present disclosure, for any variable, the variable determination module 402 can determine the discrete value corresponding to the variable, and each value (probability value) in the discrete value corresponds to a different interval range.

[0120] In one embodiment of the present disclosure, for any variable, the variable determination module 402 can determine the discrete value corresponding to the variable in the following manner: for each interval range corresponding to the variable, the following processing is performed respectively: the number of training samples in which the predetermined data corresponding to the variable is located in the interval range is counted, the ratio of the counted number of training samples to the total number of training samples is obtained, and the ratio is taken as the value corresponding to the interval range.

[0121] Further, the parameter determination module 403 can determine the network parameters of the Bayesian network model, i.e., the model parameters, according to the training samples and the values of each variable.

[0122] In one embodiment of the present disclosure, the parameter determination module 403 can pre-process the training samples, and the pre-processing can include: for any predetermined data in any training sample, the variable corresponding to the predetermined data is determined respectively, and the interval range to which the predetermined data belongs is determined from the interval range corresponding to the corresponding variable, and the value corresponding to the interval range is used to replace the predetermined data, and then the network parameters of the Bayesian network model can be determined according to the pre-processed training samples and the values of each variable.

[0123] In one embodiment of the present disclosure, the parameter determination module 403 can determine the network parameters of the Bayesian network model by maximum likelihood estimation according to the pre-processed training samples and the values of each variable.

[0124] Figure 5 The constituent structure schematic diagram of the high-precision map quality verification device embodiment 500 of the present disclosure is shown in FIG. 5. Figure 5As shown, the device comprises: a data acquisition module 501, a first determination module 502, and a second determination module 503.

[0125] The data acquisition module 501 is configured to acquire, for a high-precision map to be verified, at least one set of point cloud data pairs, the point cloud data being point cloud data used to generate the high-precision map.

[0126] The first determination module 502 is configured to, for each set of point cloud data pairs, respectively perform the following processing: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality verification result of the point cloud data pair according to the input data and network parameters of a Bayesian network model determined according to a training sample.

[0127] The second determination module 503 is configured to determine a quality verification result of the high-precision map according to the quality verification results of the point cloud data pairs.

[0128] The above-mentioned device embodiment scheme can pre-construct a Bayesian network model, and can determine network parameters of the Bayesian network model according to a training sample. Accordingly, the quality verification of the high-precision map can be automatically implemented by means of the network parameters, thereby saving manpower and time cost, improving processing efficiency, and improving the accuracy of the verification result, etc.

[0129] For each set of point cloud data pairs, the first determination module 502 can determine input data according to two frames of point cloud data therein. In an embodiment of the present disclosure, the input data can include: a data difference value corresponding to the two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, and a third intermediate result, the relative pose can include: an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value, the first intermediate result, the second intermediate result, and the third intermediate result are all intermediate results determined according to the data difference value and the relative pose between the two frames of point cloud data.

[0130] After the input data is acquired, the first determination module 502 can further determine a quality verification result of the point cloud data pair according to the input data and network parameters of a Bayesian network model determined according to a training sample.

[0131] Specifically, in an embodiment of the present disclosure, the first determination module 502 can determine a first calculation expression and a second calculation expression respectively by variable elimination method according to the Bayesian network model, then can determine a height ghosting of the point cloud data pair according to the input data, the network parameters, and the first calculation expression, and can determine a horizontal ghosting of the point cloud data pair according to the input data, the network parameters, and the second calculation expression, and further can determine the quality verification result of the point cloud data pair according to the determined height ghosting and horizontal ghosting.

[0132] Figure 4 and Figure 5 The specific working procedure of the device embodiment shown in the foregoing method embodiment will not be repeated.

[0133] In summary, the scheme described in the device embodiment of the present disclosure can automatically realize quality checking of a high-precision map, thereby saving manpower and time cost, improving processing efficiency, and improving the accuracy of the checking result.

[0134] The scheme described in the present disclosure can be applied to the field of artificial intelligence, and particularly relates to the fields of automatic driving, machine learning, and computer vision. Artificial intelligence is a discipline that studies enabling computers to simulate some thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.) of humans, and includes both hardware-level technologies and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing, etc. Artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, and machine learning / deep learning, big data processing technology, knowledge graph technology, and several other major directions.

[0135] In the technical scheme of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical scheme comply with relevant legal regulations and do not violate public order and good customs.

[0136] According to embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium, and a computer program product.

[0137] Figure 6 A schematic block diagram of an electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, servers, servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections, and their functions, as well as their implementation, are merely examples and are not intended to limit the implementations described and / or claimed herein.

[0138] As Figure 6As shown, the device 600 includes a computing unit 601 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from the storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the device 600 can also be stored in the RAM 603. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0139] Various components in the device 600 are connected to the I / O interface 605, including an input unit 606 such as a keyboard, a mouse, etc., an output unit 607 such as various types of displays, speakers, etc., a storage unit 608 such as a magnetic disk, an optical disk, etc., and a communication unit 609 such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0140] The computing unit 601 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 601 performs various methods and processes described above, such as the methods described in the present disclosure. For example, in some embodiments, the methods described in the present disclosure can be implemented as a computer software program that is tangibly embodied in a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the methods described in the present disclosure can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the methods described in the present disclosure by any other appropriate means, such as by means of firmware.

[0141] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0142] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, or entirely on a remote machine or server.

[0143] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0144] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0145] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0146] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0147] It should be understood that various forms of flow shown above can be used, with steps reordered, added, or removed. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, without limitation, as long as the desired results of the technology disclosed in the present disclosure are achieved.

[0148] The specific embodiments described above are not intended to limit the scope of the present disclosure. Those skilled in the art will understand that various modifications, combinations, sub-combinations, and alternatives can be made to the specific embodiments without departing from the spirit and principles of the present disclosure. Any further modifications, changes, improvements, and the like that come within the spirit and principles of the present disclosure should be considered within the scope of the present disclosure.

Claims

1. A model processing method, comprising: constructing a Bayesian network model; determining values of variables in the Bayesian network model according to obtained training samples, wherein each training sample corresponds to a set of point cloud data pairs, and each training sample includes the following M kinds of predetermined data: a data difference value corresponding to two frames of point cloud data in the point cloud data pairs, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, a third intermediate result, a height ghosting, and a horizontal ghosting, the relative pose including an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value, the height ghosting and the horizontal ghosting being used to reflect the quality of the point cloud data pairs, the data difference value including an absolute value of a difference between two one-dimensional variables obtained by respectively mapping point cloud data in local maps corresponding to the two frames of point cloud data to one-dimensional variables, each intermediate result being an intermediate result determined according to the data difference value and the relative pose, the first intermediate result being a ground thickness, the second intermediate result being a road edge linearity, and the third intermediate result being a wall surface flatness, and each variable in the Bayesian network model including a variable corresponding to each of the M kinds of predetermined data; determining network parameters of the Bayesian network model according to the training samples and the values of the variables, the network parameters being used to determine a quality checking result of a high-definition map to be checked, the quality checking result being determined according to input data determined based on a point cloud data pair corresponding to the high-definition map and the network parameters, and the point cloud data being point cloud data used to generate the high-definition map.

2. The method of claim 1, wherein the Bayesian network model includes a first layer of variables, a second layer of variables, and a third layer of variables; the first layer of variables includes a first variable corresponding to the data difference value, a second variable corresponding to the x-axis difference value, a third variable corresponding to the y-axis difference value, a fourth variable corresponding to the z-axis difference value, a fifth variable corresponding to the pitch angle difference value, a sixth variable corresponding to the roll angle difference value, and a seventh variable corresponding to the yaw angle difference value; the second layer of variables includes an eighth variable corresponding to the first intermediate result, a ninth variable corresponding to the second intermediate result, and a tenth variable corresponding to the third intermediate result; the third layer of variables includes an eleventh variable corresponding to the height ghosting and a twelfth variable corresponding to the horizontal ghosting.

3. The method of claim 2, wherein parent nodes of the eighth variable include the first variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable, and each variable is a node; parent nodes of the ninth variable include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable. The parent node of the tenth variable comprises the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable; The parent node of the eleventh variable comprises the eighth variable; The parent node of the twelfth variable comprises the ninth variable and the tenth variable.

4. The method of any one of claims 1-3, wherein, The determining the values of the variables in the Bayesian network model comprises: for any variable, respectively determining discrete values corresponding to the variable, each value in the discrete values corresponding to a different interval range.

5. The method of claim 4, wherein, The respectively determining the discrete values corresponding to the variable comprises: For each interval range corresponding to the variable, the following processing is respectively performed: Counting the number of training samples in which the predetermined data corresponding to the variable is located in the interval range in the training samples; Obtaining the ratio of the counted number of training samples to the total number of training samples, and taking the ratio as the value corresponding to the interval range.

6. The method of claim 5, wherein, The determining the network parameters of the Bayesian network model according to the training samples and the values of the variables comprises: Preprocessing the training samples, the preprocessing comprising: for any predetermined data in any training sample, respectively determining the variable corresponding to the predetermined data, and determining the interval range to which the predetermined data belongs from the interval range corresponding to the corresponding variable, and replacing the predetermined data with the value corresponding to the interval range to which the predetermined data belongs; Determining the network parameters of the Bayesian network model according to the preprocessed training samples and the values of the variables.

7. The method of claim 6, wherein, The determining the network parameters of the Bayesian network model according to the preprocessed training samples and the values of the variables comprises: Determining the network parameters of the Bayesian network model by maximum likelihood estimation according to the preprocessed training samples and the values of the variables.

8. A high-precision map quality verification method, comprising: For a high-precision map to be verified, obtaining at least one group of point cloud data pairs, the point cloud data being point cloud data used to generate the high-precision map; For each group of point cloud data pairs, the following processing is performed: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality check result of the point cloud data pair according to the input data and network parameters of a Bayesian network model; wherein the network parameters are determined by the method according to any one of claims 1-7; the input data includes data difference values corresponding to two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, and a third intermediate result, the relative pose includes an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value, the data difference values include absolute values of differences between two one-dimensional variables obtained by respectively mapping point cloud data in local maps corresponding to the two frames of point cloud data to one-dimensional variables, each intermediate result is an intermediate result determined according to the data difference values and the relative pose, the first intermediate result is a ground thickness, the second intermediate result is a road edge linearity, and the third intermediate result is a wall surface flatness; determining a quality check result of the high-definition map according to the quality check results of each point cloud data pair.

9. The method of claim 8, wherein, The determination of the quality check result of the point cloud data pair includes: determining a first calculation expression and a second calculation expression by variable elimination method according to the Bayesian network model; determining a height ghosting of the point cloud data pair according to the input data, the network parameters, and the first calculation expression, and determining a horizontal ghosting of the point cloud data pair according to the input data, the network parameters, and the second calculation expression; determining the quality check result of the point cloud data pair according to the height ghosting and the horizontal ghosting.

10. A model processing device, comprising: a model construction module, a variable determination module, and a parameter determination module; the model construction module is configured to construct a Bayesian network model; The variable determination module is configured to determine values of variables in the Bayesian network model according to the obtained training samples, wherein each training sample corresponds to a set of point cloud data pairs, and each training sample includes the following M types of predetermined data: a data difference value corresponding to two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result, a third intermediate result, a height ghosting, and a horizontal ghosting, the relative pose includes an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value, and a yaw angle difference value, the height ghosting and the horizontal ghosting are used to reflect the quality of the point cloud data pair, and the data difference value includes an absolute value of a difference between two one-dimensional variables obtained by respectively mapping point cloud data in local maps corresponding to the two frames of point cloud data to one-dimensional variables, each intermediate result is an intermediate result determined according to the data difference value and the relative pose, the first intermediate result is a ground thickness, the second intermediate result is a road edge linearity, and the third intermediate result is a wall surface flatness, and each variable in the Bayesian network model includes a variable corresponding to each of the M types of predetermined data. The parameter determination module is configured to determine network parameters of the Bayesian network model according to the training samples and the values of the variables, and the network parameters are used to determine a quality checking result of a high-precision map to be checked, the quality checking result is determined according to input data determined based on a point cloud data pair corresponding to the high-precision map and the network parameters, and the point cloud data is point cloud data used to generate the high-precision map.

11. The apparatus of claim 10, wherein The Bayesian network model includes a first layer of variables, a second layer of variables, and a third layer of variables. The first layer of variables includes a first variable corresponding to the data difference value, a second variable corresponding to the x-axis difference value, a third variable corresponding to the y-axis difference value, a fourth variable corresponding to the z-axis difference value, a fifth variable corresponding to the pitch angle difference value, a sixth variable corresponding to the roll angle difference value, and a seventh variable corresponding to the yaw angle difference value. The second layer of variables includes an eighth variable corresponding to the first intermediate result, a ninth variable corresponding to the second intermediate result, and a tenth variable corresponding to the third intermediate result. The third layer of variables includes an eleventh variable corresponding to the height ghosting and a twelfth variable corresponding to the horizontal ghosting.

12. The apparatus of claim 11, wherein Parent nodes of the eighth variable include the first variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable, and each variable is a node; Parent nodes of the ninth variable include the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable, and the seventh variable. The parent node of the tenth variable comprises the first variable, the second variable, the third variable, the fourth variable, the fifth variable, the sixth variable and the seventh variable; The parent node of the eleventh variable comprises the eighth variable; The parent node of the twelfth variable comprises the ninth variable and the tenth variable.

13. The apparatus of any one of claims 10-12, wherein, The variable determining module determines, for any variable, a discrete value corresponding to the variable, each value in the discrete value corresponding to a different interval range.

14. The apparatus of claim 13, wherein, The variable determining module, for each interval range corresponding to the variable, respectively performs the following processing: counts a number of training samples in which a predetermined data corresponding to the variable in the training samples is located in the interval range, obtains a ratio of the counted number of training samples to a total number of the training samples, and takes the ratio as a value corresponding to the interval range.

15. The apparatus of claim 14, wherein, The parameter determining module pre-processes the training samples, the pre-processing comprising: for any predetermined data in any training sample, respectively determining a variable corresponding to the predetermined data, and determining an interval range to which the predetermined data belongs from interval ranges corresponding to the corresponding variable, replacing the predetermined data with a value corresponding to the interval range to which the predetermined data belongs, and determining network parameters of the Bayesian network model according to the pre-processed training samples and the values of the variables.

16. The apparatus of claim 15, wherein, The parameter determining module determines network parameters of the Bayesian network model by a maximum likelihood estimation method according to the pre-processed training samples and the values of the variables.

17. A high-definition map quality verification device comprising: The data acquisition module, the first determining module and the second determining module; The data acquisition module is configured to acquire at least one group of point cloud data pairs for a high-definition map to be verified, the point cloud data being point cloud data used to generate the high-definition map. The first determination module is configured to, for each group of point cloud data pairs, perform the following processing: determining input data according to two frames of point cloud data in the point cloud data pair, and determining a quality check result of the point cloud data pair according to the input data and network parameters of a Bayesian network model; the network parameters are determined according to the method in any one of claims 1-7; the input data includes data difference values corresponding to two frames of point cloud data in the point cloud data pair, a relative pose between the two frames of point cloud data, a first intermediate result, a second intermediate result and a third intermediate result, the relative pose includes an x-axis difference value, a y-axis difference value, a z-axis difference value, a pitch angle difference value, a roll angle difference value and a yaw angle difference value, the data difference values include absolute values of difference values of two one-dimensional variables obtained by respectively mapping point cloud data in local maps corresponding to the two frames of point cloud data to one-dimensional variables, each intermediate result is an intermediate result determined according to the data difference values and the relative pose, the first intermediate result is a ground thickness, the second intermediate result is a road edge linearity, and the third intermediate result is a wall surface flatness. The second determination module is configured to determine a quality check result of the high-definition map according to the quality check results of the point cloud data pairs.

18. The apparatus of claim 17, wherein, The first determination module determines a first calculation expression and a second calculation expression by variable elimination according to the Bayesian network model, determines a height ghosting of the point cloud data pair according to the input data, the network parameters and the first calculation expression, determines a horizontal ghosting of the point cloud data pair according to the input data, the network parameters and the second calculation expression, and determines the quality check result of the point cloud data pair according to the height ghosting and the horizontal ghosting.

19. An electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

20. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-9.

21. A computer program product comprising computer programs / instructions, which, when executed by a processor, implement the method of any one of claims 1-9.

Images