A road condition information completion method based on road network topology

Abstract

The present invention discloses a method for completing road condition information based on a road network topology structure. For each determined time slice, the road condition information (speed of each road section) under the current time slice is first extracted. The road condition information of the missing road sections is completed through steps one and two using these road condition information. Finally, the road condition information of almost all road sections can be obtained. Step one: A road network topology structure is described. For the road sections that conform to this structure, a formula for completing road condition information is given. Step two: For the remaining road sections that do not conform to the described topology structure, according to their topological relationship in the road network, the largest missing subgraph of all unknown road sections is found, and then a certain step length is expanded outward according to these subgraphs. The road sections in the expanded step length are compared with the historical road sections to calculate the similarity. If the similarity is high, the data of this part of the historical road sections are used to complete the data of the missing road sections.

Description

Technical Field

[0001] The invention belongs to the field of data mining and relates to a road condition information completion method based on a road network topology structure. Background Art

[0002] Cars have become an important part of people's travel. Vehicles affect road conditions. Due to certain needs, such as traffic control, event arrangements, etc., it is very necessary to keep abreast of road conditions. Traffic condition information includes a lot of information, such as traffic flow, speed, vehicle type, etc. There are many ways to obtain traffic parameters, such as GPS positioning data, video surveillance, sensors, etc. With the widespread application of GPS positioning systems and the development of urban road networks, many researchers have studied traffic conditions through GPS data, such as estimating traffic parameters through bus GPS trajectory data. However, when using GPS data to calculate road conditions, people often cannot obtain all the positioning information of all vehicles on the road, which requires a method to complete the road condition information in order to obtain all the road condition information.

[0003] A floating vehicle data system was designed to collect information about traffic conditions on the UK's major road network through instrumented vehicles. A traffic parameter extraction method based on video surveillance can extract traffic parameters of highway sections from traffic image sequences and can accurately extract most vehicles. However, this method has defects for typical urban intersections. Videos of urban intersections generally have a mixture of pedestrians and vehicles, which reduces the effectiveness of traffic parameter extraction. A robust algorithm RTPE method can effectively extract traffic condition data at urban intersections. Some researchers use sensor data to provide the speed of track points to study real-time traffic conditions. Traffic parameters are predicted using probability models based on data from vehicles driving near traffic lights. Others have established a multivariate regression model, using bus GPS data as a probe car to analyze the average speed changes of some sections. Cycle-based traffic conditions can be used to provide intersection signal control solutions. Cycle-based traffic condition data is converted into a tensor decomposition problem, and the Tucker decomposition method is used to solve this problem.

[0004] The data used in the present invention is based on taxi GPS trajectory data, which is similar to bus GPS data. The algorithm of the present invention needs to establish a speed estimation model. Unlike previous studies, their studies mostly use multiple linear regression models, fuzzy estimation, neural network prediction, tensor decomposition and other methods, but most of them do not pay attention to the topological structure of the road network, and their methods cannot cover the entire road network. The traffic road conditions are closely related to the structure of the road network. The adjacent sections in the road network affect each other. The present invention focuses on the topological structure of the road network and uses the road condition data of the known sections to complete the road condition data of the unknown sections. Summary of the invention

[0005] The purpose of the present invention is to provide a method for completing road condition information based on a road network topology. For each determined time slice, we first extract the road condition information (speed of each road section) under the current time slice. We use this road condition information to complete the missing road condition information through steps one and two, and finally, we can obtain the road condition information of almost all road sections.

[0006] A method for completing road condition information based on a road network topology structure, the implementation steps of the method are as follows:

[0007] Step 1: We describe a road network topology as shown in the attached Figure 1 For the road segments that meet this structure, we give a formula to complete the road condition information. If the topological structure of the unknown road segment and the known road segment in the road network meets the road network topological structure we described, we use this formula to complete the road condition information of the road segment.

[0008] In the attached Figure 1 In the example, the speeds of the known sections A and C in the current time slice, and the speed of the unknown section B in the current time slice. Due to the connectivity between different sections, the traffic information between ABC is related, and B is the set of these unknown sections. Here, all unknown sections form a continuous road with no other forks. N1 links m roads, N i Links n roads.

[0009] For the attached Figure 1 This structure, assuming that the speeds of these unknown roads are v1, v2, v3, v4, ..., v i-2 ,v i-1 ,v i , we have the following formula:

[0010]

[0011] In this way, the process of solving the equation can be regarded as the process of solving a system of linear equations. By solving the system of linear equations, we can find v1,v2,v3,v4,…,v i-2 ,v i-1 ,v i , that is, we can know the traffic information of these roads.

[0012] Step 1.1: Traverse the unknown road segments from the unknown road segment set road_unknown

[0013] Step 1.2: Get the unknown road section roadid and its corresponding two nodes node1 and node2 in the road network in order

[0014] Step 1.3: Determine whether roadid has been calculated. If so, go to step 1.1. If not, go to step 1.4.

[0015] Step 1.4: For the road segment with the roadid number, determine whether it conforms to the structure of the road segment that can be completed as described. If so, go to step 1.5. If not, go to step 1.1.

[0016] Step 1.5: Expand from node1 and node2 to the surrounding areas according to the road network topology, and save the expanded road segment ID

[0017] Step 1.6: Calculate the traffic information of these expanded unknown road sections according to Formula 1.

[0018] Step 2: For the remaining road sections that do not conform to the topological structure we described, we find the largest missing subgraph of all unknown road sections based on their topological relationship in the road network, and then expand these subgraphs outward by a certain step length, and compare the road sections in the expansion step length with the historical road sections to calculate the similarity. If the similarity is high, the data of this part of the historical road sections will be used to complete the data of the missing road sections.

[0019] As attached Figure 2 As shown in Figure 1, the lines between nodes represent roads, and the nodes represent the endpoints of the roads, which together form a road network. Figure 2 The figure shows a certain area of ​​the road network under different time slices. First, we describe the road network under the current time slice (CurrentTimeSlice). Black roads indicate that the road conditions of these roads under the current time slice are unknown, and colored roads indicate that the road conditions under the current time slice are known. Due to the connectivity between roads, the road conditions of black roads are correlated with the road conditions of colored roads. We call roads with known road conditions Road known, and roads with unknown road conditions Road unknown. Then, if the road conditions of these roads under other time slices such as (TimeSlice Ti, TimeSlice Tj) are known, and the road conditions of road known in the time slice CurrentTimeSlice are similar to the road conditions of corresponding roads in other time slices, then it is very likely that the road conditions of road unknown in the time slice CurrentTimeSlice are similar to the road conditions of corresponding roads in other time slices. Based on this premise, when the road conditions of these roads are known in other time slices, we can use the road conditions of these roads in other time slices to replace the road conditions of unknown roads (road unknown) in the current time slice (CurrentTimeSlice).

[0020] The above is the ideal situation. However, in actual applications, it may occur that the known road sections of the time slices (TimeSlice Ti, TimeSlice Tj) cannot fully cover the unknown roads of the current time slice (CurrentTimeSlice). Expressed in mathematical terms, we assume that there are m1 black roads and n1 colored roads in CurrentTimeSlice. Among other time slices, there are x j roads corresponding to the black roads in CurrentTimeSlice and with known road conditions, and y j roads corresponding to the colored roads in CurrentTimeSlice and with known road conditions. The following situations may exist:

[0021]

[0022] The ideal situation is that x j = m1, y j = n1. In the ideal situation, we only need to consider the similarity between these n1 road sections in TimeSlice Tj and these n1 road sections in CurrentTimeSlice. As mentioned before, we use the average speed and speed range to represent the road conditions and define the formula:

[0023]

[0024] We can clearly see that the smaller the value of similarity, the higher the road condition similarity between CurrentTimeSlice and TimeSlice Tj.

[0025] When x j < m1, y j < n1, obviously we need to comprehensively consider the relationship between the quantity of y j and the road condition similarity. When the proportion is smaller, the error of the final result is larger. We propose a weight. For the time slice CurrentTimeSlice to be completed, the weight formula of TimeSlice Tj is as follows:

[0026]

[0027] When the proportion is larger, the difference in road conditions of these y j road sections is smaller, and the weight of TimeSlice Tj is larger, indicating that the road condition similarity between TimeSlice Tj and the missing road sections of CurrentTimeSlice is higher. And when x jWhen m1, I can use this method to find the weights after calculating other time slices and the time slice CurrentTimeSlice, and supplement the traffic conditions of the m1 missing road sections in the time slice in descending order of weights. In this way, we can obtain all the traffic conditions.

[0028] Step 2.1: Traverse the unknown road sections in the unknown road section set road_unknow in turn

[0029] Step 2.2: Obtain the unknown road section roadid and its corresponding two nodes node1 and node2 in the road network in sequence

[0030] Step 2.3: Determine whether roadid has been calculated. If so, go to Step 1.1; if not, go to Step 2.4

[0031] Step 2.4: Start from this road section to find the largest missing subgraph in this area, that is, the set of road sections that are all missing traffic conditions and are connected to each other.

[0032] Step 2.5: Expand 1 step length (represented as a road section in the road network) from the boundary nodes of this subgraph to form a new graph. The expanded road section set of the new graph is denoted as extendnode

[0033] Step 2.6: Compare the traffic conditions of extendnode with those of other time slices, and calculate the weights of different time slices according to Formula 4.

[0034] Step 2.7: Supplement the missing traffic condition information in the subgraph in descending order of weights.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1) Compared with the traditional method of using historical time slices to supplement traffic condition information, the present invention has a higher accuracy rate;

[0037] 2) Compared with the multiple linear regression method, the present invention has a higher accuracy rate;

[0038] 3) Compared with the tensor decomposition method, the present invention has higher time efficiency. Brief Description of the Drawings

[0039] Figure 1 It is a road network topology diagram of a specific structure in Step 1 of the present invention.

[0040] Figure 2 It is an explanatory diagram of the principle in Step 2.

[0041] Figure 3 It is an implementation flowchart of Step 1 of the present invention.

[0042] Figure 4This is an implementation flow chart of step 2 of the present invention. DETAILED DESCRIPTION

[0043] The present invention will be explained and described below in conjunction with the accompanying drawings:

[0044] The data set used in the present invention is the trajectory data of a taxi company in Beijing in June 2013 provided by Beijing Municipal Transportation Commission.

[0045] We first follow the Figure 3 It is executed by the steps. It should be noted that the input of the algorithm includes: the road network topology table, that is, the road numbers of the road network and the node numbers at both ends of the roads; the known road set, that is, the road numbers and road condition information; the unknown road set, that is, the unknown road numbers.

[0046] Take the execution process of the algorithm as an example:

[0047] Step 1.1 The id of the traversed road is 18974

[0048] Step 1.2: roadid = 18974, node1 = 37919, node2 = 37947

[0049] Step 1.3: The road with roadid 18974 has not been calculated, that is, the road condition is unknown, so go to step 1.4

[0050] Step 1.4: The road with roadid 18974 conforms to the specific topological structure we described, that is, there are other known roads starting from node1, and the road conditions of other nodes starting from node2 are unknown. Execute step 1.5

[0051] Step 1.5: Expand unknown roads starting from the road with roadid 18974. The expanded road ids are 18975 and 18976.

[0052] Step 1.6: Calculate the unknown road condition information according to formula 1, which is shown in Figure 1 That is, m=3, n=3, V(A)=47.00km / h, V(C)=34.72km / h, and the road set B contains 3 unknown sections. Solving the linear equations, the average speed of the road with roadid 18974 is 43.49km / h, the average speed of the road with roadid 18975 is 41.74km / h, and the average speed of the road with roadid 18976 is 39.99km / h.

[0053] Then we follow the Figure 4It is executed by the steps. It should be noted that the input of the algorithm includes: road network topology table, that is, the road number of the road network and the node number at both ends of the road; known road set, that is, road number and road condition information; unknown road set, that is, unknown road number; current time slice.

[0054] Take the execution process of the algorithm as an example:

[0055] Step 2.1 The id of the traversed road is 20523

[0056] Step 2.2: roadid = 20523, node1 = 41046, node2 = 41045

[0057] Step 2.3: The road with roadid 20523 has not been calculated, that is, the road condition is unknown, so go to step 2.4

[0058] Step 2.4: Starting from this road section, find the largest missing subgraph in this area, that is, the set of interconnected road sections with missing road conditions. The found subgraph road section set is {20523,20545,20756,20526,21203,20534,20555,20556,20733}.

[0059] Step 2.5: Expand outward from the boundary node of the subgraph by 1 step (represented as a road section in the road network) to form a new graph. The extended road section set of the new graph is extendnode = {21308, 20544, 20574}

[0060] Step 2.6: Compare the traffic conditions of extendnode with those of other time slices, and calculate the weights of different time slices according to Formula 4. The partial results are intercepted in descending order of weight values ​​as follows:

[0061] Table 1 The weight of each time slice relative to the current time slice

[0062] tweek timeslice weight 2 9 60.49298516 1 9 59.6372946 1 10 50.56027381 3 8 40.26902478 3 8 32.02341206 …… …… ……

[0063] Step 2.7: Use the data from historical time slices to complete the missing data in the subgraph in order of weight from high to low. The results are shown in Table 2.

[0064] Table 2 Road condition completion results of sub-graphs

[0065]

[0066]

Claims

1. A method for completing road condition information based on road network topology, characterized in that: The implementation steps of this method are as follows: Step 1: For a road section that conforms to the road network topology, a formula for completing the road condition information is given; if the topological structure of the unknown road section and the known road section in the road network conforms to the described road network topology, the formula is used to complete the road condition information of the road section; The speed of known sections A and C in the current time slice, and the speed of unknown section B in the current time slice; Due to the connectivity between different sections, the traffic information between ABC is related, B is the set of these unknown sections, where all unknown sections form a continuous road with no other forks; N1 links m roads respectively, N i Links n roads; For the road network topology, let the unknown speeds of these roads be: v1,v2,v3,v4,…,v i-2 ,v i-1 ,v i , we have the following formula: Find v1,v2,v3,v4,…,v i-2 ,v i-1 ,v i , that is, knowing the traffic information of these roads; Step 2: For the remaining road sections that do not conform to the topological structure described, find the largest missing subgraph of all unknown road sections according to the topological relationship in the road network, and then expand these subgraphs outward by a certain step length, and compare and calculate the similarity between the road sections in the expansion step length and the historical road sections. If the similarity is high, use the data of this part of the historical road section to complete the data of the missing road section; The lines between nodes represent roads, and the nodes represent the endpoints of roads that form a road network. First, describe the road network at the current time slice. Black roads indicate that the road conditions of these roads at the current time slice are unknown, and colored roads indicate that the road conditions at the current time slice are known. In mathematical terms, suppose there are m1 black roads and n1 colored roads in CurrentTimeSlice; in other time slices, there are x roads whose positions correspond to the black roads in CurrentTimeSlice and whose road conditions are known. j There are y roads with colored locations corresponding to CurrentTimeSlice and known road conditions. j The following situations may exist: The ideal situation is x j =m1,y j =n1. In an ideal situation, we only need to consider the similarity between the n1 road segments in the time slice TimeSlice Tj and the n1 road segments in the time slice CurrentTimeSlice. As mentioned above, the average speed and speed range are used to represent the road conditions, and the definition formula is: The smaller the similarity value is, the higher the road condition similarity between the time slice CurrentTimeSlice and the time slice TimeSlice Tj is; When x j <m1, y j <n1, it is obvious that the relationship between the quantity of y j and the similarity of road conditions needs to be comprehensively considered. When the proportion is smaller, the error of the final result is larger. A weight is proposed. For the time slice CurrentTimeSlice to be filled, the weight formula of the time slice TimeSlice Tj is as follows: When has a larger proportion, the worse the road conditions of this y j section of the road are smaller, the greater the weight of the time slice Timeslice Tj, which indicates that the road conditions similarity between the time slice TimeSlice Tj and the missing road section of the time slice CurrentTimeSlice is higher. And when x j <m1, use this method to find the weights after calculating other time slices with the time slice CurrentTimeSlice, and complete the road condition information of the m1 missing road sections of the time slice in descending order of weights; in this way, all the road condition information is obtained.

2. A method for completing road condition information based on road network topology structure according to claim 1, characterized in that: The implementation steps of step 1 of the method are as follows: Step 1.1: Traverse the unknown road sections from the unknown road section set road_unknown one by one; Step 1.2: Get the unknown road section roadid and its corresponding two nodes node1 and node2 in the road network in order; Step 1.3: Determine whether roadid has been calculated. If yes, go to step 1.

1. If no, go to step 1.

4. Step 1.4: for the road segment numbered as roadid, determine whether it conforms to the structure of the road segment that can be completed as described, if so, go to step 1.5, if not, go to step 1.1; Step 1.5: Expand from node1 and node2 to the surrounding areas according to the road network topology structure, and save the expanded road section ID; Step 1.6: Calculate the traffic information of these expanded unknown road sections according to Formula 1.

3. The method for completing road condition information based on road network topology structure according to claim 2, characterized in that: The implementation steps of step 2 of the method are as follows: Step 2.1: Traverse the unknown road sections from the unknown road section set road_unknown in sequence; Step 2.2: Obtain the unknown road section roadid and the corresponding two nodes node1 and node2 in the road network in order; Step 2.3: Determine whether roadid has been calculated. If yes, go to step 2.

1. If no, go to step 2.

4. Step 2.4: Starting from the road section, find the largest missing subgraph in the area, that is, the set of interconnected road sections with missing road conditions; Step 2.5: Expand outward from the boundary node of the subgraph by one step, which is represented as a road segment in the road network to form a new graph; the set of extended road segments in the new graph is recorded as extendnode; Step 2.6: Compare the traffic condition of extendnode with the traffic condition of other time slices, and calculate the weights of different time slices according to formula 4; Step 2.7: Complete the missing traffic information in the subgraph in descending order of weight.

Images