Start point binding method and device, computer equipment and storage medium

Abstract

The invention discloses a starting point binding method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining historical navigation information of a target user and a current starting point position; obtaining candidate roads in a preset range of the current starting point position, and extracting starting point features, candidate road features and cross road features between the current starting point position and the candidate roads; determining the joint probability of each road based on the current starting point position; determining Boolean features of each candidate road based on the historical navigation information; and determining a target starting point road based on the starting point feature, the candidate road feature, the cross road feature, the joint probability of each road and the Boolean feature. Historical navigation information and preorder track information of a user are effectively utilized, multiple features are adopted to jointly screen candidate roads, a target starting point road is obtained, starting point road binding is achieved by combining multi-factor information, accuracy can be improved, subjective dependence can be reduced, and the application scene is wider.

Description

Technical Field

[0001] The present application relates to the field of navigation technology, and in particular to a method, device, computer equipment and storage medium for binding a starting point. Background Art

[0002] In navigation applications, when a user initiates a navigation request, it is crucial to accurately know the road where the user's current starting point is located. Only by accurately binding the starting point to a specific road on the road network can a reasonable and accurate driving route be planned for the user in combination with the destination and real-time traffic information. If the starting point is bound to a road that is not properly bound, the entire route planned subsequently may be wrong and fail to meet the user's needs to reach the destination efficiently and accurately.

[0003] At present, the starting point binding method usually uses the road topological relationship and attribute information in the map data to match the trajectory starting point with a road with similar topological characteristics on the road network. For example, based on the number of road intersections around the starting point, the curvature of the road, the road grade and other information, the road that best matches these features is found as the starting point binding road. The accuracy of this method depends largely on the quality of the map data. The definition and matching criteria of road features may be subjective, have certain limitations, and have low accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a starting point binding method, device, computer equipment and storage medium to solve at least one problem existing in the above-mentioned prior art.

[0005] In a first aspect, a method for binding a starting point is provided, comprising:

[0006] Obtain the target user's historical navigation information and current starting point location;

[0007] Acquire candidate roads within a preset range of the current starting point position, and extract starting point features, candidate road features, and intersection road features between the current starting point position and the candidate roads;

[0008] Based on the current starting point position, determining the joint probability of each road;

[0009] Determining Boolean features of each candidate road based on the historical navigation information;

[0010] Based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature, a target starting point road is determined.

[0011] In one embodiment, after obtaining the historical navigation information and the current starting point location of the target user, the method further includes:

[0012] Get the actual trajectory points of the target user within the preset time range;

[0013] Determine the binding point of the actual trajectory point in the road network, and use the binding point as a true value road to verify the target starting point road based on the true value road.

[0014] In one embodiment, determining the binding point of the actual trajectory point in the road network includes:

[0015] Determine a projection point of the actual trajectory point in the road network, use the projection point as a hidden point, and use the actual trajectory point as an observation point;

[0016] Determine the observation probability of each hidden point to the observation point, and the transition probability between a hidden point and the next hidden point;

[0017] Based on the observation probability, the transition probability and the joint probability of the last hidden point, obtaining the joint probability of each hidden point;

[0018] The hidden point with the largest joint probability is used as the binding point corresponding to the actual trajectory point.

[0019] In one embodiment, determining the joint probability of each road based on the current starting point position includes:

[0020] Based on the current starting point position, determining a preceding point;

[0021] Determine the corresponding joint probability between the preceding point and surrounding roads.

[0022] In one embodiment, determining the corresponding joint probability between the preceding point and surrounding roads includes:

[0023] Determine the binding point corresponding to the preceding point in the road network;

[0024] Determine the joint probability between the binding point corresponding to the last preceding point in the preceding points and the surrounding roads to serve as the joint probability of the roads.

[0025] In one embodiment, determining the Boolean features of each candidate road based on the historical navigation information includes:

[0026] Based on the historical navigation information, determining the corresponding end point position when the last navigation ended;

[0027] Determine the distance between the end point and each candidate road;

[0028] A Boolean feature corresponding to each candidate road sign is generated based on the distance value.

[0029] In one embodiment, determining the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature includes:

[0030] Perform feature splicing on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature to obtain a spliced ​​feature;

[0031] Inputting the splicing features into a pre-trained model to score each candidate road based on the pre-trained model;

[0032] Based on the scoring results, the candidate roads are sorted in ascending or descending order;

[0033] Based on the ranking results, the candidate road with the highest score is used as the target starting point road.

[0034] In a second aspect, a starting point binding device includes:

[0035] A navigation request acquisition unit, used to acquire the historical navigation information and current starting point location of the target user;

[0036] A feature extraction unit, configured to obtain candidate roads within a preset range of the current starting point position, and extract starting point features, candidate road features, and cross-road features between the current starting point position and the candidate roads;

[0037] A joint probability determination unit, configured to determine the joint probability of each road based on the current starting point position;

[0038] A Boolean feature determination unit, configured to determine a Boolean feature of each candidate road based on the historical navigation information;

[0039] The target starting road determination unit is used to determine the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature.

[0040] In a third aspect, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and running on the processor, wherein the processor implements the above-mentioned starting point binding method when executing the computer-readable instructions.

[0041] In a fourth aspect, a readable storage medium is provided, on which computer-readable instructions are stored, wherein when the computer-readable instructions are executed by a processor, the starting point binding method as described above is implemented.

[0042] The above-mentioned starting point binding method, device, computer equipment and storage medium, the method implementation includes: obtaining the historical navigation information and current starting point position of the target user; obtaining the candidate roads within the preset range of the current starting point position, and extracting the starting point features, candidate road features and the cross-road features between the current starting point position and the candidate roads; determining the joint probability of each road based on the current starting point position; determining the Boolean features of each candidate road based on the historical navigation information; determining the target starting point road based on the starting point features, candidate road features, cross-road features, joint probabilities of each road and the Boolean features. In the embodiment of the present application, the historical navigation information and the preceding trajectory information of the user are effectively utilized, and the obtained joint probabilities of each road, Boolean features and road features of the candidate roads around the starting point, the starting point features of the current starting point position and the cross-road features between the starting point and the candidate roads are used together to screen the candidate roads to obtain the target starting point road, and the starting point binding is realized by combining multi-factor information, which can improve accuracy, reduce subjective dependence, and have a wider range of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0044] Figure 1 This is a flow chart of a method for binding a starting point in an embodiment of the present application;

[0045] Figure 2 This is a schematic diagram of the structure of a starting point binding device in one embodiment of the present application;

[0046] Figure 3 It is a schematic diagram of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] In one embodiment, if Figure 1 As shown, a starting point binding method is provided, comprising the following steps:

[0049] In step S110, the historical navigation information and current starting point location of the target user are obtained;

[0050] In the embodiment of the present application, historical navigation information can be obtained by querying the request log, and the historical navigation information can specifically include the last navigation end point, the current navigation starting point, navigation time and other information. The current starting point can be a specific road name, longitude and latitude information, etc., which can be obtained through GPS or other positioning devices. Optionally, when a navigation request is sent by a target user, the historical navigation information of the target user can be obtained and the current starting point of the target user can be located.

[0051] In step S120, candidate roads within a preset range of the current starting point are obtained, and starting point features, candidate road features, and cross-road features between the current starting point and the candidate roads are extracted;

[0052] Optionally, after obtaining the current starting position, possible candidate roads can be recalled based on distance. For example, the current starting position can be used as the center of the circle, a preset radius value, such as 1000 meters, can be set, and then all road segments and road nodes connected to the circular area can be screened from the road network database, and these roads can be used as candidate roads. Alternatively, possible candidate roads can be recalled based on a graph data structure. For example, the road network can be abstracted into a graph data structure, in which the intersections or endpoints of the roads are used as nodes and the road segments are used as edges. Starting from the node corresponding to the starting position, the roads corresponding to the directly connected neighbor nodes and the edges connecting these neighbor nodes are searched, and these roads are the most basic candidate roads. Then, the roads corresponding to the neighbor nodes of the neighbor nodes can be further recursively searched, and the range of the recalled roads can be gradually expanded until certain conditions are met, for example, the preset search depth is reached or the number of recalled roads reaches a certain threshold, at which time all candidate roads can be obtained. All candidate roads recalled through the current starting position can be stored in the candidate road set for subsequent use.

[0053] The starting point features may specifically include the latitude and longitude, speed, direction, accuracy and other features of the GPS point.

[0054] The candidate road features may specifically include the road shape angle, road attributes, whether tolls are charged, road width and other features of each candidate road.

[0055] The intersection features may include information such as the distance and direction angle between the starting point and each candidate road.

[0056] In step S130, based on the current starting point position, the joint probability of each road is determined;

[0057] Optionally, the previous point information can be obtained based on the current starting point position. For example, the travel trajectory data of the target user can be obtained, and the trajectory points before the starting point in chronological order can be used as the previous points. Alternatively, the historical navigation information recorded in the map application can be combined to find the location points that the user passed before the starting point position as the previous points. Then, the joint probability between each previous point and the roads around each previous point is calculated using the hidden Markov algorithm to serve as the joint probability of each road.

[0058] The predecessor point refers to the relevant information of other points before a certain point in the user trajectory sequence. This information can include various information that can describe the state of the trajectory point, such as location (such as longitude and latitude), timestamp, speed, direction, etc. For example, in a user trajectory containing multiple trajectory points A, B, C, and D, for point C, the information of points A and B is its predecessor point information.

[0059] In step S140, based on the historical navigation information, a Boolean feature of each candidate road is determined;

[0060] Optionally, the historical navigation information may include navigation information such as the end position and time of the last navigation. The navigation information such as the end position and time of the last navigation may be matched with each candidate road respectively, and the matching result is used as the Boolean feature of the candidate road recalled this time. The Boolean feature is used to indicate whether the candidate road is the end position of the last navigation.

[0061] In step S150, a target starting road is determined based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road, and the Boolean feature.

[0062] Optionally, a sorting model can be constructed, and the starting point features, candidate road features, intersection road features, joint probabilities of each road, and the Boolean features can be input into the sorting model, and each candidate road can be scored and sorted in ascending or descending order of the score to obtain a sorting result, and based on the sorting result, the candidate road with the highest score is determined as the target starting point road.

[0063] The ranking model may be a lambdaMART ranking model, which, as a combination of lambda-rank and decision tree, can further improve the ranking performance by integrating multiple gradient boosting tree models. This integration method can be achieved by increasing the number of trees or using different feature subsets. It should be noted that other ranking models may also be used, such as RankNet, LightGBM, etc.

[0064] In an embodiment of the present application, a method for binding a route to a starting point is provided, including: obtaining the historical navigation information and the current starting point position of the target user; obtaining candidate roads within a preset range of the current starting point position, and extracting starting point features, candidate road features, and cross-road features between the current starting point position and the candidate roads; determining the joint probability of each road based on the current starting point position; determining the Boolean features of each candidate road based on the historical navigation information; and determining the target starting point road based on the starting point features, candidate road features, cross-road features, joint probabilities of each road, and the Boolean features. In an embodiment of the present application, the historical navigation information and the preceding trajectory information of the user are effectively utilized, and the obtained joint probabilities of each road, Boolean features, and road features of the candidate roads around the starting point, the starting point features of the current starting point position, and the cross-road features between the starting point and the candidate roads are used together to screen the candidate roads to obtain the target starting point road. Combining multi-factor information to realize starting point binding can improve accuracy, reduce subjective dependence, and have a wider range of application scenarios.

[0065] In an embodiment of the present application, after obtaining the navigation request of the target user, the method further includes:

[0066] Based on the navigation request, obtaining actual track points of the target user within a preset time range;

[0067] Determine the binding point of the actual trajectory point in the road network, and use the binding point as a true value road to verify the target starting point road based on the true value road.

[0068] Optionally, the actual trajectory point can obtain the coordinate information of the target user's location at different times through a positioning device such as GPS. The projection point of the actual trajectory point on the road network can be used as a hidden point, and the actual trajectory point can be used as an observation point. Then, the observation probability of each hidden point to the observation point and the transfer probability from the hidden point to the next hidden point are calculated, and then the joint probability is calculated through the joint probability calculation formula. Finally, the hidden point with the largest joint probability is determined and used as the binding point of the previous point on the road network. After completing the binding operation of all actual trajectory points, the binding points of each actual trajectory point on the road network can be obtained.

[0069] Then, the binding point can be used as the true value road to verify the target starting road based on the true value road. For example, verify whether the target starting road is consistent with the true value road. If they are consistent, it means that the prediction is accurate. Otherwise, the model needs to be adjusted.

[0070] In an embodiment of the present application, determining the binding point of the actual trajectory point in the road network includes:

[0071] Determine a projection point of the actual trajectory point in the road network, use the projection point as a hidden point, and use the actual trajectory point as an observation point;

[0072] Determine the observation probability of each hidden point to the observation point, and the transition probability between a hidden point and the next hidden point;

[0073] Based on the observation probability, the transition probability and the joint probability of the last hidden point, obtaining the joint probability of each hidden point;

[0074] The hidden point with the largest joint probability is used as the binding point corresponding to the actual trajectory point.

[0075] Optionally, the distance from the actual trajectory point to each road segment on the road network (for example, the road can be regarded as consisting of multiple segments) can be calculated, and then a suitable position on the road segment closest to the trajectory point is determined as the projection point. For example, for a trajectory point, the vertical distance to several surrounding roads is calculated (or other suitable distance measurement methods), and if the distance to a certain road is the shortest, the point closest to the trajectory point on this road is used as the projection point. Exemplarily, the actual trajectory point of the target user shows that it is located in a certain block in a certain city. After processing, it is determined that its projection point in the road network is at the intersection where a main road and a branch road intersect. This intersection position is the projection point for the actual trajectory point.

[0076] Then, the projection point can be used as a hidden point and the actual trajectory point as an observation point. For each hidden point (that is, a potential location point on each road in the road network), it is necessary to calculate its observation probability to the observation point and the transition probability from the hidden point to the next hidden point. The observation probability is used to measure the probability of the current observation trajectory point appearing under a given hidden point hypothesis. The transition probability reflects the possibility of transferring from one hidden point position to the next hidden point position in the road network, such as considering factors such as road connectivity and traffic rules. Then, the joint probability of each hidden point is calculated in turn through the joint probability calculation formula, that is, joint probability = observation probability × transition probability × joint probability of the previous hidden point. Finally, each hidden point with the largest joint probability is determined and used as the binding point of the actual trajectory point on the road network. The binding operation of all actual trajectory points is completed to obtain the binding points of each actual trajectory point on the road network.

[0077] In an embodiment of the present application, determining the joint probability of each road based on the current starting point position includes:

[0078] Based on the current starting point position, determining a preceding point;

[0079] Determine the corresponding joint probability between the preceding point and surrounding roads.

[0080] It can be understood that the predecessor point refers to the relevant information of other points before a certain point in the user trajectory sequence. This information can include various information that can describe the state of the trajectory point, such as location (such as longitude and latitude), timestamp, speed, direction, etc. For example, in a user trajectory containing multiple trajectory points A, B, C, and D, for point C, the information of points A and B is its predecessor point information.

[0081] Optionally, the predecessor point information can be obtained based on the current starting point position. For example, the travel trajectory data of the target user can be obtained, and the trajectory points before the starting point in chronological order can be used as the predecessor points. Alternatively, the historical navigation information recorded in the map application can be combined to find the location points that the user passed before the starting point as the predecessor points. Then, the joint probability between each predecessor point and the roads around each predecessor point is calculated according to the hidden Markov algorithm as the joint probability of each road. Exemplarily, a point sequence of length 25 can be maintained in real time before the target user sets off to store the user's pre-navigation information, that is, the predecessor point information. When navigation is initiated, the hidden Markov algorithm can be used based on these 25 points to calculate the joint probability of all roads around the last point.

[0082] In an embodiment of the present application, determining the corresponding joint probability between the preceding point and surrounding roads includes:

[0083] Determine the binding point corresponding to the preceding point in the road network;

[0084] Determine the joint probability between the binding point corresponding to the last preceding point in the preceding points and the surrounding roads to serve as the joint probability of the roads.

[0085] Optionally, the distance from the previous point to each road segment on the road network (for example, the road can be regarded as composed of multiple segments) can be calculated, and then the position on the road segment closest to the previous point can be determined as the projection point. For example, for a previous point, the vertical distances from it to several surrounding roads are calculated respectively. If the distance to a certain road is the shortest, then the point on this road closest to the trajectory point is used as the projection point. Exemplarily, the previous point of the target user shows that it is located in a certain block in a certain city. After processing, it is determined that its projection point in the road network is at the intersection where a main road and a branch road intersect. This intersection position can be used as the projection point of the previous point.

[0086] Then, the projection point can be used as a hidden point, and the previous point can be used as an observation point. For each hidden point (that is, a potential location point on each road in the road network), it is necessary to calculate its observation probability to the observation point and the transition probability from the hidden point to the next hidden point. The observation probability is used to measure the probability of the current observed previous point appearing under a given hidden point hypothesis; the transition probability reflects the possibility of transferring from one hidden point position to the next hidden point position in the road network, such as considering factors such as road connectivity and traffic rules. Then, the joint probability of each hidden point is calculated in turn through the joint probability calculation formula, that is, joint probability = observation probability × transition probability × joint probability of the previous hidden point. For example, if there are 25 previous points, after completing the entire calculation process based on 25 points, focus on the situation corresponding to the last point. At this time, for the last point, the joint probability of all roads around it (that is, the roads where the corresponding hidden points are located) will be obtained. Finally, each hidden point with the largest joint probability is determined, and it is used as the binding point of the previous point on the road network. The binding operation of all previous points is completed to obtain the binding point of each previous point on the road network.

[0087] The joint probability quantitatively represents the possibility of the last point appearing on each road. For example, the joint probability corresponding to a main road is 0.3, and the joint probability corresponding to a branch road is 0.1, etc. It describes the degree of correlation between the last point and the surrounding roads from a probability perspective.

[0088] In one embodiment, determining the Boolean features of each candidate road based on the historical navigation information includes:

[0089] Based on the historical navigation information, determining the corresponding end point position when the last navigation ended;

[0090] Determine the distance between the end point and each candidate road;

[0091] A Boolean feature corresponding to each candidate road sign is generated based on the distance value.

[0092] It is understandable that the historical navigation information may include key information of each navigation process of the target user, such as starting point information, route, end location, etc. By screening and extracting the historical navigation information, the corresponding end location and navigation end time at the end of the last navigation can be accurately located.

[0093] Optionally, after obtaining the end point position, the distance value between the end point position and each candidate road can be calculated, for example, by using the Euclidean distance algorithm. For example, if the coordinates of the end point position are (x1, y1) and the coordinates of a point on the candidate road, such as the midpoint of a road segment, are (x2, y2), then the distance Alternatively, you can also use the road network shortest path distance calculation, such as Dijkstra algorithm or A algorithm, etc., with the end point as the starting point, search the road network for the shortest path length to a reference point on each candidate road (such as a road entry point, the road node closest to the end point, etc.), and use it as the distance value between the two.

[0094] It should be noted that the Boolean feature (flag_lastnav) has only two values, true or false. If the distance value is smaller, it means that the road is the end position of the last navigation, then the road is marked as true, and other roads are marked as false.

[0095] In an embodiment of the present application, determining the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature includes:

[0096] Perform feature splicing on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature to obtain a spliced ​​feature;

[0097] Inputting the splicing features into a pre-trained model to score each candidate road based on the pre-trained model;

[0098] Based on the scoring results, the candidate roads are sorted in ascending or descending order;

[0099] Based on the ranking results, the candidate road with the highest score is used as the target starting point road.

[0100] Optionally, the starting point features, candidate road features, cross road features, joint probabilities of each road, and the Boolean features can be spliced ​​to obtain spliced ​​features. Then, the spliced ​​features are input into the pre-trained model, scored and sorted by the pre-trained model, and then the candidate road with the highest score can be used as the target starting point road. The pre-trained model can be a lambdaMART sorting model, which, as a combination of lambda-rank and decision tree, can further improve the sorting performance by integrating multiple gradient boosting tree models. This integration method can be achieved by increasing the number of trees or using different feature subsets.

[0101] It should be noted that all the features corresponding to each request log can be combined into a query. In a query, the road with the same ID as the true road is scored as 1, and other roads are scored as 0. The data is divided into training set, validation set, and test set in a ratio of 7:2:1. First, the training set data is feature cleaned to filter out the cases where the true road binding fails, the starting speed direction is illegal, and the candidate roads do not contain the true road. The lambdaMART model is based on a decision tree, and the features can be directly spliced ​​and then transmitted to the model.

[0102] Use the validation set data to verify the model effect. Input the validation set features into the model, take the road with the highest score in each query as the final road, and verify the consistency between the output road and the true road as the accuracy.

[0103] The training process of the model can be shown in Table 1 below:

[0104] Table 1: Pseudo code table of lambdaMART training process

[0105]

[0106] When the candidate roads are sorted, a random sorting result can be obtained. If the order of road i and road j is swapped, the DCG (discount gain) of the entire sorting list will change, and deltaDCGij = deltaDCGji is obtained, that is, the absolute value of the difference in DCG obtained by swapping the positions of the two roads.

[0107] Therefore, this value reflects the impact of exchanging the two options. If the impact is large, it means that the grade (label) of the two roads is very different, otherwise the difference is small.

[0108] This leads to the definition of lambda(i,j) for i and j:

[0109] λ ij =ρ ij *|deltaNDCG ij |;

[0110] The coefficient ρ can be understood as the probability that i is worse than j, that is, if road i is better than road j, then the value of ρ is small, and the impact of swapping the positions of the two should be small. The above is the lambda calculation method for document pair ij. The most important thing is the lambda of document i, which is not the cumulative sum, but the following formula:

[0111]

[0112] Then, lambda is used as the label and the concatenated features are used as input to train the decision tree. During the training process, the goal is to minimize the ranking loss. For example, the ranking loss function can be defined as L = 1-NDCG p , The ideal maximum DCG (IDCG) is the DCG value when all relevant documents are ranked from high to low in terms of relevance.

[0113] This method can calculate the lambda of the originally non-differentiable NDCG indicator so that the gradient can be calculated normally, and the decision tree model itself will not have a gradient explosion.

[0114] In the embodiment of the present application, the user's historical navigation information and previous trajectory information are effectively used to screen the candidate roads by combining the joint probability of each road, the Boolean features, the road features of the candidate roads around the starting point, the starting point features of the current starting point position, and the cross-road features between the starting point and the candidate roads to obtain the target starting point road. The starting point binding is achieved by combining multi-factor information, which can improve accuracy, reduce subjective dependence, and have a wider range of application scenarios. Experiments have shown that the above method can enable the model to achieve better sorting effects. Compared with manual strategies, the accuracy is improved by 10%. Compared with the sorting algorithm that does not use historical trajectory information and navigation information, the accuracy is improved by 3.6%.

[0115] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0116] In one embodiment, a starting point binding device is provided, and the starting point binding device corresponds one-to-one with the starting point binding method in the above embodiment. Figure 2 As shown, the starting point road binding device includes a navigation request acquisition unit 10, a feature extraction unit 20, a joint probability determination unit 30, a Boolean feature determination unit 40 and a target starting point road determination unit 50. The functional modules are described in detail as follows:

[0117] A navigation request acquisition unit 10 is used to acquire the historical navigation information and the current starting point location of the target user;

[0118] A feature extraction unit 20 is used to obtain candidate roads within a preset range of the current starting point position, and extract starting point features, candidate road features, and cross-road features between the current starting point position and the candidate roads;

[0119] A joint probability determination unit 30, configured to determine the joint probability of each road based on the current starting point position;

[0120] A Boolean feature determination unit 40, configured to determine the Boolean feature of each candidate road based on the historical navigation information;

[0121] The target starting road determining unit 50 is used to determine the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature.

[0122] In one embodiment of the present application, the device further includes a true value road determination unit, which is used to:

[0123] Get the actual trajectory points of the target user within the preset time range;

[0124] Determine the binding point of the actual trajectory point in the road network, and use the binding point as a true value road to verify the target starting point road based on the true value road.

[0125] In one embodiment of the present application, the true value road determination unit is further used to:

[0126] Determine a projection point of the actual trajectory point in the road network, use the projection point as a hidden point, and use the actual trajectory point as an observation point;

[0127] Determine the observation probability of each hidden point to the observation point, and the transition probability between a hidden point and the next hidden point;

[0128] Based on the observation probability, the transition probability and the joint probability of the last hidden point, obtaining the joint probability of each hidden point;

[0129] The hidden point with the largest joint probability is used as the binding point corresponding to the actual trajectory point.

[0130] In one embodiment of the present application, the joint probability determination unit 30 is further configured to:

[0131] Based on the current starting point position, determining a preceding point;

[0132] Determine the corresponding joint probability between the preceding point and surrounding roads.

[0133] In one embodiment of the present application, the joint probability determination unit 30 is further configured to:

[0134] Determine the binding point corresponding to the preceding point in the road network;

[0135] Determine the joint probability between the binding point corresponding to the last preceding point in the preceding points and the surrounding roads to serve as the joint probability of the roads.

[0136] In one embodiment of the present application, the Boolean feature determination unit 40 is further configured to:

[0137] Based on the historical navigation information, determining the corresponding end point position when the last navigation ended;

[0138] Determine the distance between the end point and each candidate road;

[0139] A Boolean feature corresponding to each candidate road sign is generated based on the distance value.

[0140] In one embodiment of the present application, the target starting road determination unit 50 is further configured to:

[0141] Perform feature splicing on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature to obtain a spliced ​​feature;

[0142] Inputting the splicing features into a pre-trained model to score each candidate road based on the pre-trained model;

[0143] Based on the scoring results, the candidate roads are sorted in ascending or descending order;

[0144] Based on the ranking results, the candidate road with the highest score is used as the target starting point road.

[0145] In the embodiment of the present application, the user's historical navigation information and previous trajectory information are effectively utilized, and the joint probability of each road, the Boolean features and the road features of the candidate roads around the starting point, the starting point features of the current starting point position, and the intersection road features between the starting point and the candidate roads are jointly used to screen the candidate roads to obtain the target starting point road. Combining multi-factor information to achieve starting point binding can improve accuracy, reduce subjective dependence, and have a wider range of application scenarios.

[0146] For the specific definition of the starting point binding device, please refer to the definition of the starting point binding method above, which will not be repeated here. Each module in the above starting point binding device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0147] In one embodiment, a computer device is provided. The computer device may be a terminal device, and its internal structure diagram may be as follows: Figure 3As shown. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a readable storage medium. The readable storage medium stores computer-readable instructions. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer-readable instructions are executed by the processor, a starting point binding method is implemented. The readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.

[0148] In an embodiment of the present application, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the steps of the starting point binding method described above are implemented.

[0149] In an embodiment of the application, a readable storage medium is provided, wherein the readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the above-mentioned starting point binding method are implemented.

[0150] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through computer-readable instructions, and the computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When the computer-readable instructions are executed, they may include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0151] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0152] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A starting point binding method, characterized in that: The method comprises: Obtain the target user's historical navigation information and current starting point location; Acquire candidate roads within a preset range of the current starting point position, and extract starting point features, candidate road features, and intersection road features between the current starting point position and the candidate roads; Based on the current starting point position, determining the joint probability of each road; Determining Boolean features of each candidate road based on the historical navigation information; Based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature, a target starting point road is determined.

2. The starting point binding method according to claim 1, characterized in that: After obtaining the historical navigation information and the current starting point location of the target user, the method further includes: Get the actual trajectory points of the target user within the preset time range; Determine the binding point of the actual trajectory point in the road network, and use the binding point as a true value road to verify the target starting point road based on the true value road.

3. The starting point binding method according to claim 2, characterized in that: The determining of the binding point of the actual trajectory point in the road network comprises: Determine a projection point of the actual trajectory point in the road network, use the projection point as a hidden point, and use the actual trajectory point as an observation point; Determine the observation probability of each hidden point to the observation point, and the transition probability between a hidden point and the next hidden point; Based on the observation probability, the transition probability and the joint probability of the last hidden point, obtaining the joint probability of each hidden point; The hidden point with the largest joint probability is used as the binding point corresponding to the actual trajectory point.

4. The starting point binding method according to claim 1, characterized in that: The determining of the joint probability of each road based on the current starting point position includes: Based on the current starting point position, determining a preceding point; Determine the corresponding joint probability between the preceding point and surrounding roads.

5. The starting point binding method according to claim 4, characterized in that: The determining of the corresponding joint probability between the preceding point and surrounding roads includes: Determine the binding point corresponding to the preceding point in the road network; Determine the joint probability between the binding point corresponding to the last preceding point in the preceding points and the surrounding roads to serve as the joint probability of the roads.

6. The starting point binding method according to claim 1, characterized in that: The determining of the Boolean features of each candidate road based on the historical navigation information includes: Based on the historical navigation information, determining the corresponding end point position when the last navigation ended; Determine the distance between the end point and each candidate road; A Boolean feature corresponding to each candidate road sign is generated based on the distance value.

7. The starting point binding method according to claim 1, characterized in that: The determining of the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature includes: Perform feature splicing on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature to obtain a spliced ​​feature; Inputting the splicing features into a pre-trained model to score each candidate road based on the pre-trained model; Based on the scoring results, the candidate roads are sorted in ascending or descending order; Based on the ranking results, the candidate road with the highest score is used as the target starting point road.

8. A starting point binding device, characterized in that: The device comprises: A navigation request acquisition unit, used to acquire the historical navigation information and current starting point location of the target user; A feature extraction unit, configured to obtain candidate roads within a preset range of the current starting point position, and extract starting point features, candidate road features, and cross-road features between the current starting point position and the candidate roads; A joint probability determination unit, configured to determine the joint probability of each road based on the current starting point position; A Boolean feature determination unit, configured to determine a Boolean feature of each candidate road based on the historical navigation information; The target starting road determination unit is used to determine the target starting road based on the starting point feature, the candidate road feature, the intersection road feature, the joint probability of each road and the Boolean feature.

9. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executed on the processor, characterized in that: When the processor executes the computer-readable instructions, the starting point binding method according to any one of claims 1 to 7 is implemented.

10. A readable storage medium having computer readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by a processor, the starting point binding method according to any one of claims 1 to 7 is implemented.

Images