Map area matching method and device

Abstract

The invention discloses a map area matching method and device, and the method comprises the steps: obtaining first map data and second map data corresponding to a to-be-matched area, and obtaining a first path data set and a second path data set according to an entering path and a leaving path of the first map data and the second map data; respectively taking the first path data set and the second path data set as a state space and an action space, and executing reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data; and determining a regional road network matching relationship between the first map data and the second map data according to the two groups of state-action value data obtained by reinforcement learning. According to the method, the matching problem of different source maps can be effectively solved, the calculation complexity is low, the method can adapt to standard and format differences of different map data, the problem of cross-map standard inconsistency is relieved, and a foundation is laid for automatic fusion of different map data in subsequent work.

Description

Technical Field

[0001] The present invention relates to the technical field of map processing, and particularly to a method and device for map area matching. Background Art

[0002] With the continuous development of autonomous driving technology, it is necessary to effectively integrate map data from multiple different sources, including traditional standard definition map data from different manufacturers and new high-precision map data. Traditional map data provides basic geographical information, while high-precision maps provide more detailed and accurate data, including complex road geometries and traffic signs. By integrating these two types of data, the autonomous driving system can better adapt to diverse driving scenarios, increase redundancy and fault tolerance, and obtain real-time updated map information, thereby improving performance and reliability in various complex environments. This fusion strategy aims to provide a more comprehensive, accurate, and real-time map foundation for autonomous driving vehicles to support their safe and efficient navigation decisions. When fusing two different map data sets, ensuring accurate matching between different map areas is crucial. In the prior art, feature matching algorithms are mainly used for map area matching. Summary of the Invention

[0003] To obtain more accurate map area matching results, embodiments of the present invention provide a method and device for map area matching.

[0004] In a first aspect, embodiments of the present invention provide a method for map area matching, the method comprising:

[0005] Obtain first map data and second map data corresponding to an area to be matched, and respectively obtain a first route data set and a second route data set according to the incoming route and the outgoing route of the first map data and the second map data;

[0006] Respectively use the first route data set and the second route data set as the state space and the action space, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data;

[0007] Determine the area road network matching relationship between the first map data and the second map data according to the two sets of state-action value data obtained by reinforcement learning.

[0008] In one or some optional embodiments of the embodiments of the present application, the step of respectively using the first route data set and the second route data set as the state space and the action space, and performing reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data includes:

[0009] Take the first route data set as the first state space and the second route data set as the first action space;

[0010] Based on the preset action-value function, calculate the first initial state-action value data;

[0011] Based on the preset reinforcement learning algorithm, iteratively update the first initial state-action value data until a preset condition is reached to obtain the first state-action value data;

[0012] And,

[0013] Take the first route data set as the second action space and the second route data set as the second state space;

[0014] Based on the preset action-value function, calculate the second initial state-action value data;

[0015] Based on the preset reinforcement learning algorithm, iteratively update the second initial state-action value data until a preset condition is reached to obtain the second state-action value data.

[0016] In one or some alternative embodiments of the embodiments of the present application, the determining the regional road network matching relationship between the first map data and the second map data according to the two sets of state-action value data obtained by reinforcement learning includes:

[0017] Respectively determine the first matching relationship and the second matching relationship between the first route data set and the second route data set according to the maximum values of each row of the first state-action value data and the second state-action value data;

[0018] Respectively determine whether the first matching relationship and the second matching relationship are valid matching relationships:

[0019] If both the first matching relationship and the second matching relationship are valid matching relationships and the first matching relationship is the same as the second matching relationship, then obtain the regional road network matching relationship;

[0020] If both the first matching relationship and the second matching relationship are invalid matching relationships, then re-perform reinforcement learning;

[0021] If only one of the first matching relationship and the second matching relationship is a valid matching relationship, then determine whether the valid matching relationship meets the prior knowledge condition;

[0022] If it meets the prior knowledge condition, then use the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship;

[0023] If the prior knowledge condition is not met, reinforcement learning is re-executed.

[0024] In one or some alternative embodiments of the present application, if the first matching relationship or the second matching relationship meets the prior knowledge condition, after using the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship, the following steps are further included:

[0025] Determine the data difference between the first state-action value data or the second state-action value data corresponding to the invalid matching relationship, determine the dividing line through the gradient magnitude between adjacent values, and obtain the verified regional road network matching relationship;

[0026] If the regional road network matching relationship meets the verified regional road network matching relationship, it is determined that the regional road network matching relationship is credible.

[0027] In one or some alternative embodiments of the present application, the following steps are further included:

[0028] If the regional road network matching relationship does not meet the verified regional road network matching relationship, determine whether the difference between two adjacent values corresponding to the maximum gradient of the verified data difference exceeds a preset threshold. If so, it is determined that the regional road network matching relationship is credible.

[0029] In one or some alternative embodiments of the present application, after obtaining the regional road network matching relationship, the following steps are further included:

[0030] According to the regional road network matching relationship, move the images corresponding to the first map data and the second map data, calculate the coincidence degree of all matching routes in the two images after the movement, and determine whether the coincidence degree meets a preset coincidence degree threshold;

[0031] If so, it is determined that the regional road network matching relationship is credible.

[0032] In one or some alternative embodiments of the present application, calculating the first initial state-action value data based on the preset action value function includes:

[0033] Based on the preset action value function, calculate the first initial state-action value corresponding to each route in the first state space and each route in the first action space, and construct the first initial state-action value data;

[0034] And / or,

[0035] Calculating the second initial state-action value data based on the preset action value function includes:

[0036] Based on the preset action value function, calculate the second initial state-action values corresponding to each path in the second state space and each path in the second action space, and construct the second initial state-action value data.

[0037] In one or some optional embodiments of the embodiments of the present application, the step of using the first path data set as the first state space and using the second path data set as the first action space includes:

[0038] Partition the first path data set, map at least one path to one state to obtain the first state space, and / or partition the second path data set, map at least one path to one action to obtain the first action space;

[0039] and / or

[0040] The step of using the first path data set as the second action space and using the second path data set as the second state space includes:

[0041] Partition the first path data set, map at least one path to one action to obtain the second action space, and / or partition the second path data set, map at least one path to one state to obtain the second state space.

[0042] In one or some optional embodiments of the embodiments of the present application, the step of obtaining the first map data and the second map data corresponding to the area to be matched, and respectively obtaining the first path data set and the second path data set according to the entry paths and exit paths of the first map data and the second map data includes:

[0043] Obtain the first map data and the second map data corresponding to the area to be matched, and determine the entry paths and exit paths of the first map data and the second map data;

[0044] Based on the entry paths and exit paths of the first map data and the second map data, perform path partitioning to obtain the first path data set and the second path data set.

[0045] In a second aspect, an embodiment of the present invention provides a map area matching device, which includes:

[0046] An acquisition module, configured to acquire the first map data and the second map data corresponding to the area to be matched, and respectively obtain the first path data set and the second path data set according to the entry paths and exit paths of the first map data and the second map data;

[0047] A reinforcement learning module, configured to use the first route data set and the second route data set as a state space and an action space respectively, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data;

[0048] A matching module, configured to determine a regional road network matching relationship between the first map data and the second map data according to two sets of state-action value data obtained by reinforcement learning.

[0049] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the map area matching method as described above is implemented.

[0050] In a fourth aspect, an embodiment of the present invention provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the map area matching method as described above is implemented.

[0051] In a fifth aspect, an embodiment of the present invention provides a computer program product containing instructions, and when the computer program product runs on a computer device, the computer device is enabled to execute the map area matching method as described above.

[0052] In a sixth aspect, an embodiment of the present invention provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is configured to run a computer program or instruction to implement the map area matching method as described above.

[0053] The beneficial effects of the above technical solutions provided by the embodiments of the present invention at least include:

[0054] The map area matching method provided by the embodiment of the present invention respectively obtains corresponding route data sets according to two map data to be matched, then uses the two route data sets as a state space and an action space respectively, and based on a reinforcement learning algorithm, obtains two sets of state-action value data to obtain a regional road network matching relationship. It can effectively solve the matching problem of different source maps, has a lower computational complexity, can adapt to the differences in standards and formats of different map data, alleviates the problem of cross-map standard inconsistency, lays a foundation for the subsequent work to realize the automatic fusion of different map data, and provides accurate and complete map information.

[0055] Other features and advantages of the present invention will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the written specification, claims, and drawings.

[0056] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0057] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention. In the accompanying drawings:

[0058] Figure 1 It is a schematic diagram of the steps of the map area matching method provided by the embodiment of the present invention;

[0059] Figure 2 It is a schematic diagram of the first map data provided by the embodiment of the present invention;

[0060] Figure 3 It is a schematic diagram of the second map data provided by the embodiment of the present invention;

[0061] Figure 4 It is a schematic diagram of the overlap of the first map data and the second map data provided by the embodiment of the present invention;

[0062] Figure 5 It is a schematic diagram of one of the first route data sets provided by the embodiment of the present invention;

[0063] Figure 6 It is a schematic diagram of another of the first route data sets provided by the embodiment of the present invention;

[0064] Figure 7 It is a schematic diagram of yet another of the first route data sets provided by the embodiment of the present invention;

[0065] Figure 8 It is a schematic diagram of still another of the first route data sets provided by the embodiment of the present invention;

[0066] Figure 9 It is a schematic diagram of one of the first route data sets provided by the embodiment of the present invention;

[0067] Figure 10 It is a schematic diagram of one of the second route data sets provided by the embodiment of the present invention;

[0068] Figure 11 It is a schematic diagram of another of the second route data sets provided by the embodiment of the present invention;

[0069] Figure 12 It is a schematic diagram of the structure of the map area matching device provided by the embodiment of the present application. Detailed Embodiments

[0070] In the following description, specific details such as specific system architectures and technologies are presented for purposes of illustration and not limitation, so as to provide a thorough understanding of the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from obscuring the description of the present application.

[0071] It should be understood that when used in the specification and appended claims of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0072] It should also be understood that the term "and / or" as used in the specification and appended claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0073] As used in the specification and appended claims of the present application, the term "if" can be interpreted as "when" or "once" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted as meaning "once determined" or "in response to determining" or "once detected [the described condition or event]" or "in response to detecting [the described condition or event]" depending on the context.

[0074] In addition, in the description of the specification and appended claims of the present application, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0075] Reference to "an embodiment" or "some embodiments" or the like described in the specification of the present application means that a specific feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, statements such as "in an embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily all refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "comprising", "including", "having", and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0076] It should be understood that the sequence numbers of the steps in the following embodiments do not indicate the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0077] To illustrate the technical solution of the present application, specific embodiments will be used for illustration below.

[0078] The inventor found that in the prior art, there is a problem of inconsistent standard formats between map data from different sources (such as standard definition map data and high-precision map data), and the data in the same map area may have obvious offsets. Based on this, the inventor made further research and developed the present invention to provide a map area matching method and device.

[0079] Embodiment 1

[0080] The embodiment of the present invention provides a map area matching method. Referring to Figure 1 as shown, the method includes:

[0081] S101: Obtain first map data and second map data corresponding to the area to be matched, and respectively obtain a first route data set and a second route data set according to the incoming routes and outgoing routes of the first map data and the second map data.

[0082] In the embodiment of the present application, first, two map data from different sources corresponding to the area to be matched are obtained, namely the first map data and the second map data. Then, according to the road network topological relationship between the first map data and the second map data, the incoming routes and outgoing routes of the first map data and the second map data are determined. Based on the incoming routes and outgoing routes of the first map data and the second map data, path division is performed to obtain a first route data set and a second route data set.

[0083] Among them, after obtaining the first map data and the second map data, since map data from different sources may cause certain offsets in the data, methods such as key point matching by extracting obvious feature points (such as landmarks, intersections, special road shapes, etc.) from the map data can be used to ensure that the first map data and the second map data belong to the same map area. The specific implementation process can refer to the detailed description of the prior art and will not be elaborated here.

[0084] S102: Respectively use the first route data set and the second route data set as the state space and the action space, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data.

[0085] In the embodiment of the present application, in the above step S102, the first path dataset and the second path dataset are respectively used as the state space and the action space, and reinforcement learning is performed based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data, which specifically may include:

[0086] First, the first path dataset is used as the first state space, and the second path dataset is used as the first action space. That is, the first path dataset is divided, and at least one path is mapped to a state to obtain the first state space. At the same time, the second path dataset is divided, and at least one path is mapped to an action to obtain the first action space.

[0087] Second, based on the preset action value function, the first initial state-action value data is calculated. Based on the preset action value function, the first initial state-action value corresponding to each path in the first state space and each path in the first action space is calculated, and the first initial state-action value data is constructed. Among them, the preset action value function can use similarity detection algorithms such as GraphMatch. Through GraphMatch, the matching score of two paths is calculated, and the matching score is used as the initial state-action value of the two paths.

[0088] Third, the first initial state-action value data is iteratively updated based on the preset reinforcement learning algorithm until a preset condition is reached to obtain the first state-action value data. Specifically, in the present application, the preset reinforcement learning algorithm can be selected according to actual needs. A value-based reinforcement learning algorithm such as Q-learning or DQN (Deep Q Network) can be used, and the corresponding update rule and preset condition are obtained according to the selected reinforcement learning algorithm.

[0089] Taking the Q-Learning algorithm as an example, its update rule is the Bellman equation. Then, in the process of iteratively updating the first initial state-action value data based on Q-Learning to obtain the first state-action value data, the first initial state-action value data will be updated based on the Bellman equation. Among them, the Bellman equation formula is as follows:

[0090] Q(S,A)=(1-α)Q(S,A)+α(r+β(Q(S′,A)-Q(S,A))) Formula 1;

[0091] In the formula, Q(S, A) represents the state-action value corresponding to state S and action A, S′ represents the next moment state, that is, the new state after executing the action, α represents the learning rate, r represents the immediate reward for executing action A, that is, the matching score of the two paths corresponding to state S and action A is calculated using the preset action value function, and β represents the discount factor.

[0092] Taking the Q-Learning algorithm as an example, the preset condition for stopping the iterative update of the first initial state-action value data is shown in Formula 2:

[0093] α(r + β(Q(S′, A) - Q(S, A))) < preset threshold, Formula 2;

[0094] Fourthly, use the first path dataset as the second action space and the second path dataset as the second state space. That is, divide the first path dataset, map at least one path to an action to obtain the second action space. At the same time, divide the second path dataset, map at least one path to a state to obtain the second state space.

[0095] Fifthly, calculate the second initial state-action value data based on the preset action value function. Based on the preset action value function, calculate the second initial state-action value corresponding to each path in the second state space and each path in the second action space, and construct the second initial state-action value data.

[0096] Sixthly, iteratively update the second initial state-action value data based on the preset reinforcement learning algorithm until the preset condition is reached to obtain the second state-action value data. The specific implementation of this step is similar to the third step above and will not be elaborated here.

[0097] In the embodiments of the present application, through the above first to sixth steps, two rounds of reinforcement learning are completed based on the preset reinforcement learning algorithm to obtain the first state-action value data and the second state-action value data. The above steps from the first to the sixth step can be executed in sequence. It is also possible to execute the first to the third step and the fourth to the sixth step in parallel to save program and time.

[0098] In the embodiments of the present application, by using the two path datasets as the action space and the state space respectively and optimizing based on the reinforcement learning algorithm, the association rules and feature relationships between the two path datasets are included in the state-action value data, which helps to understand the similarities and differences between different data in the dataset. At the same time, this method of swapping the action space and the state space and performing the second round of reinforcement learning helps to more comprehensively understand the relationship between the two path datasets, facilitating subsequent comprehensive consideration by combining the two state-action value data.

[0099] S103: Determine the regional road network matching relationship between the first map data and the second map data according to the two sets of state-action value data obtained by reinforcement learning.

[0100] In the embodiments of the present application, in the above step S103, to determine the regional road network matching relationship between the first map data and the second map data according to two sets of state-action value data obtained by reinforcement learning, it includes:

[0101] Respectively determine the first matching relationship and the second matching relationship between the first route data set and the second route data set according to the maximum values of each row of the first state-action value data and the second state-action value data;

[0102] First, respectively determine whether the first matching relationship and the second matching relationship are valid matching relationships:

[0103] In the first case, if both the first matching relationship and the second matching relationship are valid matching relationships, and the first matching relationship is the same as the second matching relationship, then use the first matching relationship as the regional road network matching relationship.

[0104] In the second case, if both the first matching relationship and the second matching relationship are invalid matching relationships, then re-execute the reinforcement learning.

[0105] In the third case, if only one of the first matching relationship and the second matching relationship is a valid matching relationship, then determine whether the valid matching relationship meets the prior knowledge condition;

[0106] If it meets the prior knowledge condition, then use the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship;

[0107] If it does not meet the prior knowledge condition, then re-execute the reinforcement learning.

[0108] In the embodiments of the present application, if the matching relationship obtained according to the first state-action value data or the second state-action value data is a uniquely determined matching result, it can be used as a valid matching relationship. If there may be two or more matching results for the matching relationship obtained according to the first state-action value data or the second state-action value data, then determine that the matching relationship is an invalid matching relationship.

[0109] In a specific embodiment, in the above third case, when only one of the first matching relationship and the second matching relationship is a valid matching relationship, and the valid matching relationship meets the prior knowledge condition, after using the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship, the method may further include:

[0110] For the invalid matching relationships among the first matching relationship and the second matching relationship, determine the data difference of the first state-action value data or the second state-action value data corresponding to the invalid matching relationship, determine the dividing line through the gradient magnitude between adjacent values, and obtain the verification regional road network matching relationship;

[0111] At this time, if the regional road network matching relationship is consistent with the regional road network matching relationship for verification, the regional road network matching relationship is determined to be credible; if the regional road network matching relationship does not meet the regional road network matching relationship for verification, it is determined whether the difference between the two adjacent values of the corresponding maximum gradient of the verification data difference exceeds the preset threshold. If so, the regional road network matching relationship is determined to be credible, otherwise information requiring manual judgment is returned.

[0112] In the embodiment of the present application, the above-mentioned prior knowledge condition may be an expected regional road network matching relationship predetermined based on the first map data and the second map data before executing the present solution or during the execution of the present solution.

[0113] In the embodiment of the present application, after obtaining the regional road network matching relationship, in order to verify whether the regional road network matching relationship is credible, the image corresponding to the first map data and the second map data can also be moved according to the regional road network matching relationship, and the overlap of all matching paths in the two images after the movement is calculated, and it is determined whether the overlap meets the preset overlap threshold. If so, it is determined that the regional road network matching relationship is credible. Among them, the method for calculating the overlap can use existing technologies such as directly calculating the overlap (Intersection over Union, IoU) and intersection over union ratio (Overlap Ratio) of the images corresponding to the first map data and the second map data, determining the image centroid and calculating the Euclidean distance of key points, etc. The specific implementation process can refer to the detailed description of the prior art, which will not be repeated here. Those skilled in the art can reasonably set the size of the above-mentioned preset overlap threshold according to actual use requirements.

[0114] In an embodiment of the present application, after determining the regional road network matching relationship, the first map data and the second map data can be fused according to the regional road network matching relationship to obtain fused regional map data. The specific method for implementing the topological relationship fusion of the two map data can refer to the detailed description of the prior art and is not specifically limited here. By fusing the two map data, data expansion of map data from different sources can be achieved. For example, by matching high-precision map data to standard-definition map data, lane-level (display) navigation in the standard-definition map can be achieved, and by matching standard-definition map data to high-definition map data, the atlas relationship in the high-definition map can be corrected.

[0115] In order to facilitate those skilled in the art to understand the present solution, the specific implementation process of the map area matching method provided in the embodiment of the present invention will be described by way of example below:

[0116] The first map data and the second map data corresponding to the to-be-matched area are as follows: Figure 2 , Figure 3As shown, it can be seen that the first map data contains 5 routes, and the second map data contains 2 routes. The schematic diagram of the direct overlap between the first map data and the second map data is as Figure 4 shown. It can be seen that there is a certain deviation between the first map data and the second map data, and it is impossible to directly determine the corresponding matching relationship between the 5 routes in the first map data and the 2 routes in the second map data. Therefore, through the map area matching method provided by the embodiments of the present invention, the corresponding matching relationship between the 5 routes in the first map data and the 2 routes in the second map data can be obtained, realizing the regional road network matching of the first map data and the second map data.

[0117] The first route dataset and the second route dataset obtained based on the first map data and the second map data are as Figures 5 to 11 shown, where Figures 5 to 9 is the first route dataset, which are the 5 routes from the inside to the outside in the first map data, Figure 10 and Figure 11 is the second route dataset, which are the inner route and the outer route in the second map data respectively.

[0118] Taking the first route dataset as the state space, each route is mapped to a state, and taking the second route dataset as the action space, each route is mapped to an action. Based on the preset action value function GraphMatch algorithm, the matching score of any state and action in the state space and the action space is calculated to obtain the first initial state-action value data, as shown in Table 1. Among them, HD_0 to HD_4 respectively represent the five states in the state space, and SD_0 and SD_1 represent the two actions in the action space. Based on the preset reinforcement learning algorithm Q-Learning, the first initial state-action value data is iteratively updated to obtain the first state-action value data as shown in Table 2.

[0119] Table 1 First initial state-action value data

[0120] HD_0 HD_1 HD_2 HD_3 HD_4 SD_0 0.4130 0.6343 0.5300 0.3922 0.4534 SD_1 0.4048 0.4255 0.3722 0.4791 0.4940

[0121] Table 2 First state-action value data

[0122] SD_0 SD_1 HD_0 2.46685955 2.45263368 HD_1 2.47383582 2.44743241 HD_2 2.56243363 2.41773241 HD_3 2.44463577 2.52334280 HD_4 2.49943955 2.53860922

[0123] Similarly, taking the second meridian dataset as the state space, each meridian is mapped to a state, and taking the first meridian dataset as the action space, each meridian path is mapped to an action. Among them, HD_0 to HD_4 represent five actions in the action space, and SD_0 and SD_1 represent two states in the state space. Based on the preset action-value function GraphMatch algorithm and the preset reinforcement learning algorithm Q-Learning, the second state-action value data is obtained, as shown in Table 3.

[0124] Table 3 Second state-action value data

[0125] HD_0 HD_1 HD_2 HD_3 HD_4 SD_0 2.83842462 2.85974613 2.75541687 2.61768791 2.67881024 SD_1 2.69257778 2.71324962 2.65994125 2.76681802 2.78175778

[0126] According to the maximum values in each row of the first state-action value data and the second state-action value data, the first matching relationship and the second matching relationship are determined. Referring to Table 2, the actions where the maximum values in each row of the first state-action value data are located are SD_0, SD_0, SD_0, SD_1, SD_1 in sequence. The first matching relationship is that HD_0, HD_1, HD_2 match SD_0, and HD_3 and HD_4 match SD_1. Referring to Table 3, the action where the maximum value in the first row of the second state-action value data is located is HD_1, and the action where the maximum value in the second row is located is HD_4. It is impossible to determine whether the dividing line should be between HD_0 - HD_1 or between HD_3 - HD_4. The obtained second matching relationship includes two or more possible matching results. Therefore, the second matching relationship is an invalid matching relationship.

[0127] At this time, only the first matching relationship is a valid matching relationship among the first matching relationship and the second matching relationship. And in this example, the prior knowledge condition is that HD_0, HD_1, HD_2 match SD_0, and HD_3 and HD_4 match SD_1, which is the same as the first matching relationship. Therefore, the first matching relationship is taken as the regional road network matching relationship.

[0128] Meanwhile, based on the second state-action value data corresponding to the second matching relationship of the invalid matching relationship, the difference between adjacent two state-action values is calculated along the row with a larger dimension, and the data difference is obtained, as shown in Table 4. The dividing line is determined by the magnitude of adjacent values in Table 4. Among them, the maximum value in each row is the difference between HD_2 and HD_3. It can be determined that the dividing line for road network matching should be between HD_2 and HD_3. Thus, the verification regional road network matching relationship obtained is that HD_0, HD_1, HD_2 match SD_0, and HD_3 and HD_4 match SD_1. This verification regional road network matching relationship is the same as the obtained regional road network matching relationship (the first matching relationship). Therefore, it can be determined that this regional road network matching relationship is credible.

[0129] Data difference of the second matching relationship in Table 4

[0130] HD_0 - HD_1 HD_1 - HD_2 HD_2 - HD_3 HD_3 - HD_4 SD_0 0.02132151 0.10432926 0.13772896 0.06112233 SD_1 0.02067184 0.05330837 0.10687677 0.01493976

[0131] In this example, since the matching relationship of the regional road network for verification is the same as the obtained matching relationship of the regional road network (the first matching relationship), it is determined that the matching relationship of the regional road network is credible. Therefore, based on the matching relationship of the regional road network, the fusion of the first map data and the second map data can be achieved to obtain the fused regional map data.

[0132] In the embodiments of the present application, reasonable decisions are made according to the different situations of two sets of state-action value data to obtain the matching relationship of the regional road network, and the credibility of the matching relationship of the regional road network is judged. The matching problem of different source maps is effectively solved, the computational complexity is relatively low, it can adapt to the standards and format differences of different map data, the problem of cross-map standard inconsistency is alleviated, and it lays a foundation for the subsequent work to realize the automatic fusion of different map data, and provides accurate and complete map information.

[0133] Embodiment 2

[0134] Based on the same inventive concept, the embodiments of the present application also provide a map area matching device. Referring to Figure 12 as shown, the device includes:

[0135] An acquisition module 101, configured to acquire the first map data and the second map data corresponding to the area to be matched, and respectively obtain a first route data set and a second route data set according to the incoming route and the outgoing route of the first map data and the second map data;

[0136] A reinforcement learning module 102, configured to respectively use the first route data set and the second route data set as the state space and the action space, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data;

[0137] A matching module 103, configured to determine the regional road network matching relationship between the first map data and the second map data according to two sets of state-action value data obtained by reinforcement learning.

[0138] Embodiment 3

[0139] Based on the same inventive concept, the embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the map area matching method described in Embodiment 1 above.

[0140] Embodiment 4

[0141] Based on the same inventive concept, an embodiment of the present application further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the map area matching method described in the first embodiment above is implemented.

[0142] Embodiment Five

[0143] Based on the same inventive concept, an embodiment of the present application further provides a computer program product containing instructions. When the computer program product runs on a computer device, the computer device is caused to execute the map area matching method described in the first embodiment above.

[0144] Embodiment Six

[0145] Based on the same inventive concept, an embodiment of the present application further provides a chip. The chip includes a processor and a communication interface. The communication interface is coupled to the processor, and the processor is configured to run a computer program or instructions to implement the map area matching method described in the first embodiment above.

[0146] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0147] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0148] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one or more flows and / or blocks Figure 1The functions specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the steps of the functions specified in Figure 1 one process or more processes and / or boxes Figure 1 the functions specified in one box or more boxes.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for map area matching, characterized in that, Including: Obtain the first map data and the second map data corresponding to the area to be matched, and respectively obtain a first route dataset and a second route dataset according to the entry routes and exit routes of the first map data and the second map data; Respectively use the first route dataset and the second route dataset as the state space and the action space, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data; Determine the regional road network matching relationship between the first map data and the second map data according to the two sets of state-action value data obtained by reinforcement learning.

2. The method according to claim 1, wherein The step of respectively using the first route dataset and the second route dataset as the state space and the action space, and performing reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data includes: Use the first route dataset as the first state space and the second route dataset as the first action space; Based on the preset action value function, calculate the first initial state-action value data; Iteratively update the first initial state-action value data based on the preset reinforcement learning algorithm until a preset condition is reached to obtain the first state-action value data; And Use the first route dataset as the second action space and the second route dataset as the second state space; Based on the preset action value function, calculate the second initial state-action value data; Iteratively update the second initial state-action value data based on the preset reinforcement learning algorithm until a preset condition is reached to obtain the second state-action value data.

3. The method according to claim 2, wherein The step of determining the regional road network matching relationship between the first map data and the second map data according to the two sets of state-action value data obtained by reinforcement learning includes: Respectively determine a first matching relationship and a second matching relationship between the first route dataset and the second route dataset according to the maximum values of each row of the first state-action value data and the second state-action value data; Respectively determine whether the first matching relationship and the second matching relationship are valid matching relationships: If both the first matching relationship and the second matching relationship are valid matching relationships and the first matching relationship is the same as the second matching relationship, then obtain the regional road network matching relationship; If both the first matching relationship and the second matching relationship are invalid matching relationships, then re-perform reinforcement learning; If only one of the first matching relationship and the second matching relationship is a valid matching relationship, then determine whether the valid matching relationship meets the prior knowledge condition; If it meets the prior knowledge condition, then use the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship; If it does not meet the prior knowledge condition, then re-perform reinforcement learning.

4. The method according to claim 3, wherein After using the first matching relationship or the second matching relationship that meets the prior knowledge condition as the regional road network matching relationship, it further includes: Determine the data difference between the first state-action value data or the second state-action value data corresponding to the invalid matching relationship, determine the dividing line based on the gradient magnitude between adjacent values, and obtain the verification regional road network matching relationship; If the regional road network matching relationship conforms to the verification regional road network matching relationship, determine that the regional road network matching relationship is credible.

5. The method according to claim 4, wherein It further includes: If the regional road network matching relationship does not conform to the verification regional road network matching relationship, determine whether the difference between two adjacent values corresponding to the maximum gradient of the verification data difference exceeds a preset threshold. If so, determine that the regional road network matching relationship is credible.

6. The method according to any one of claims 3 to 5, characterized in that, After obtaining the regional road network matching relationship, it further includes: According to the regional road network matching relationship, move the images corresponding to the first map data and the second map data, calculate the coincidence degree of all matching routes in the two images after the movement, and determine whether the coincidence degree conforms to a preset coincidence degree threshold; If so, determine that the regional road network matching relationship is credible.

7. The method according to claim 2, wherein The calculation of the first initial state-action value data based on the preset action value function includes: Based on the preset action value function, calculate the first initial state-action value corresponding to each route in the first state space and each route in the first action space, and construct the first initial state-action value data; and / or The calculation of the second initial state-action value data based on the preset action value function includes: Based on the preset action value function, calculate the second initial state-action value corresponding to each route in the second state space and each route in the second action space, and construct the second initial state-action value data.

8. The method according to claim 2, wherein The act of taking the first route dataset as the first state space and taking the second route dataset as the first action space includes: Partition the first route dataset, map at least one route to a state to obtain the first state space, and / or partition the second route dataset, map at least one route to an action to obtain the first action space; and / or The act of taking the first route dataset as the second action space and taking the second route dataset as the second state space includes: Partition the first route dataset, map at least one route to an action to obtain the second action space, and / or partition the second route dataset, map at least one route to a state to obtain the second state space.

9. The method according to claim 1, characterized in that, The act of obtaining the first map data and the second map data corresponding to the area to be matched, and respectively obtaining the first route dataset and the second route dataset according to the incoming routes and outgoing routes of the first map data and the second map data includes: Obtain the first map data and the second map data corresponding to the area to be matched, and determine the incoming routes and outgoing routes of the first map data and the second map data; Based on each entry route and each exit route of the first map data and the second map data, perform route division to obtain a first route dataset and a second route dataset.

10. A map area matching device, characterized in that, Including: An acquisition module, configured to acquire first map data and second map data corresponding to a region to be matched, and respectively obtain a first route dataset and a second route dataset according to the entry routes and exit routes of the first map data and the second map data; A reinforcement learning module, configured to respectively use the first route dataset and the second route dataset as a state space and an action space, and perform reinforcement learning based on a preset action value function and a preset reinforcement learning algorithm to obtain corresponding state-action value data; A matching module, configured to determine the regional road network matching relationship between the first map data and the second map data according to two sets of state-action value data obtained by reinforcement learning.

11. A computer-readable storage medium, in which instructions are stored. When the instructions are run on a terminal, the terminal is caused to execute the map region matching method according to any one of claims 1-9.

12. A computer device, characterized in that, Including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the map region matching method according to any one of claims 1-11 is implemented.

13. A computer program product containing instructions. When the computer program product runs on a computer device, the computer device is caused to execute the map region matching method according to any one of claims 1-11.

Images