JP2000322402A - Stream of people analyzing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a traffic line analyzing method capable of spatial analysis in addition to time analysis by determining a desired area to investigate the stream of people, selecting a spot/road to be concerned in that area, preparing a graph regarding the spot as a nodal point/regarding the road as a branch, collecting the profiles and trajectory data of respective monitors to a trajectory collection server and analyzing the collected trajectory data. SOLUTION: A structure of a street is expressed in a graph structure and the structure of a route and the value of the route are expressed by adjacent matrixes. The collected trajectory data of people are stored in an action pattern table 1 and a profile table 2. A mouth-to-mouth information exchange model 3 mutually exchanges information between persons having the information, repeats evaluation and extracts a feature amount from the information of the street. For example, a geographical condition is added to a genetic algorithm and processed. Concerning the characteristics of an action pattern, the route of real walking is expressed as the arrangement of crossing numbers and this is inherited in action. Namely, concerning two action patterns, the gene manipulation of crossing and mutation is performed and the alternation of generation of a gene is progressed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for analyzing a flow of a person based on information on a trajectory of a person moving. In particular, the present invention relates to a technique for analyzing a tendency of the entire flow based on a part of trajectory information sampled.

[0002]

2. Description of the Related Art Conventional techniques related to the present invention, such as a positioning method for measuring the position of a person, a method for acquiring flow information, a method for analyzing a person's flow, and a display method, will be described.

(1) Positioning method For example, PHS (Personal Handyph)
Using a wireless telephone system such as one system, it is possible to efficiently perform position search and the like. PHS
In the control center connected to the mobile telephone network, the location registration information of the mobile telephone terminal is analyzed based on the input search information, and the position of the specific terminal can be known.

As another positioning technique, GPS (Globa) is used.
l Positioning System).
GPS is a positioning satellite system that detects weak radio waves from artificial satellites and detects positions such as latitude and longitude, and is a system developed by the United States Department of Defense. Since the test was released for consumer use, the range of applications has expanded to not only applications such as ship navigation where positioning is indispensable, but also business use such as truck dispatch management and taxi operation management, and personal uses such as car navigation systems. ing.

[0005] Demonstration tests are also being conducted on a positioning method in which a position detection marker is arranged on a sidewalk or a wall surface and position information from the marker is received. Literature "Keisuke Suwa, et al .:" Proposal of Personal Positioning System for Pedestrian Navigation ", Proc. SA of ITE (Intelligent Transport Systems) Information and Communication Technology Symposium
According to D-6-8, pp. 125-126 (1998), the positioning accuracy of this marker method can be increased to 50 mm.

(2) Method of Obtaining Person Flow Information On the other hand, there are the following three types of existing methods for obtaining a person flow as data.

[0007] Fixed point observation method (eg: traffic flow survey device / Japanese Patent Laid-Open No. 7-23)
No. 7835) ・ Entrance / exit survey method (eg: public transportation) ・ Follow-up survey method (eg: position search system / Japanese Patent Application Laid-Open No.
No. 47974) In the prior art, for example, if a tracking method is used,
Although it was possible to track the position of a person having a GPS or the like for some people, it was not possible to acquire flow information of people throughout the city. In the fixed point observation method and entrance / exit survey method,
When the path of the person is fixed, the macro flow of the person can be obtained, but in the case of the complicated path, it cannot be investigated.

(3) Human Flow Analysis Method For the human flow analysis technology using the positioning technology, for example, “Position search system, position search method, information collection system, information collection method and in-vehicle communication device” (Japanese Patent Laid-Open No. -4
No. 7974). This system uses PHS (Pe
It is an object of the present invention to efficiently perform a position search and the like by using a wireless telephone system such as an rsonal Handyphone System. A control center connected to a mobile telephone network analyzes location registration information of a mobile telephone terminal based on input search information, searches for a specific terminal position, and outputs the search result. . According to this technology, based on location information from PHS terminals, a time-series change in the number of terminals registered in each base station or a flow of a person at a specific position is time-series as a change in the number of terminals. Can be shown.

[0009] Each terminal registers its location with one base station. The location of the base station is a relatively wide area composed of a plurality of base stations. If it is determined that two areas are adjacent to each other on a map when the terminal moves, many people use this area. Can be estimated, and this data can be used to grasp the flow of people in time series.

The Huff model, which can roughly grasp the flow of people, is a trade area analysis model often used in area marketing. According to Ryuichi Kunitomo, “Understandable Convenience Store Industry” and Nihon Jitsugyo Shuppan (1997), a map system based on the rule (Huff model) based on the rule that “the attractiveness of a store is inversely proportional to the distance to the store” is used. Therefore, the trade area is analyzed by expressing the trade area in a circular shape. This model is used for planning a franchise chain store, but there is a problem that it will not work when a competitor opens a store in a commercial area.

For example, when a convenience store is located, even stores of the same series may be intentionally located on the opposite side of a road or in the immediate vicinity of an existing store. There are a large number of stores in urban areas, but they are being opened one after another in places that only seem to be saturated in terms of Huff Model. In the case of stores that “stop by” like a convenience store, the location conditions are fundamentally different between the east and west sides of the road, but the Huff Model could not express a detailed flow of people.

(4) Person Flow Display Method In the prior art of the person flow display method, a method of expressing a trade area in a circular shape like a Huff model, a method of tracking and displaying a specific person's action route, and a method of displaying a specific flow of a person. In many cases, the output format was a time-based statistical graph of the number of people passing through the place. As described above, these methods are insufficient in expressing the flow of a person because the information that can be obtained is limited and the analysis method is simple.

[0013]

[Problems to be Solved by the Invention] "Position search system,
The position search method, the information collection system, the information collection method, and the on-vehicle communication device are technically equivalent to observing the flow of people at a fixed point, and can perform only the same analysis as a street survey using a counter. Was. To analyze the flow of people,
Spatial analysis of the route from which the person came is indispensable.

An object of the present invention is to collect the trajectory information obtained by continuously acquiring the position information of a person and analyze the trajectory information to output the tendency of the flow of the person in more detail. .

[0015] The flow of people referred to here is information on the number of people passing through a certain route, time distribution, and the like. As a result of the analysis, for example, “people who pass station 1 to station 2
0.9%. , And "60.6% of people in a residential area pass (or have passed) on this road."

[0016]

The above object is achieved by the following constitution. In a preferred embodiment of the present invention, the mobile terminal includes a mobile terminal that transmits a moved trajectory, a trajectory collection server that collects trajectory information, and an analysis terminal. Further, a flow line analysis method for analyzing the collected trajectory data is stored in the server. This flow line analysis method includes the following five steps.

The first step is to determine an area in which the flow of a person is to be investigated, select a point / road of interest, and create a graph in which the point is regarded as a node / road as a branch. Age, etc.), a second step of collecting trajectory data in the trajectory collection server, and a third step of associating the latitude and longitude data of the trajectory collected by the monitor with the nodes and branches of the graph.
, A fourth step of analyzing trajectory data, and a fifth step of extracting features by statistical processing and outputting the extracted data to the analysis terminal.

According to the above-described means, it is possible to provide a flow line analysis method capable of performing spatial analysis in addition to temporal analysis, and the object of the present invention is achieved.

Here, in the analysis processing of the fourth step, the following word-of-mouth information exchange model is used.

Word-of-mouth information exchange model / step 1: Each person learns the information of his walking path.

Word-of-mouth information exchange model / step 2: Meet friends who have walked in the same city. Or read information magazines.

Word-of-mouth information exchange model / step 3: Information exchange is performed for the routes that have been walked by each other. For interesting routes, we will incorporate information from our partners and use them the next time we visit the city. Based on the new information, partially replaced by the previous route, walk around the city and rate the new route. In addition, as a result of the information exchange, portions that are found to be following the same route with each other have an increased evaluation value. However, even if you incorporate information, you may not go on a whim, or you may adopt a completely new route due to some environmental change (such as opening a new store).

Word-of-mouth information exchange model / step 4: Some people may not come to this town due to environmental changes. Conversely, a person who has never been here may be newly visited.

Word-of-mouth information exchange model / step 5: Returning to step 2, information exchange is repeated, and everyone seeks a better walking method. As a result, there are several busy roads and places where people are structurally easy to gather.

By this repetitive processing, excellent behavior information survives natural selection, and the evaluation value of the path of the behavior pattern increases. If it repeats for the expected number of times, the data is totaled and analyzed for each route and intersection. Since this result is, so to speak, a simulation of the flow of people in the entire city, the flow of people is directly displayed on a map from time to time at the analysis terminal. In places where there are many people, visual representations of how many people actually move on the screen are displayed. In addition, for a specific route or intersection, the characteristics of the pedestrian there are analyzed.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a diagram showing the overall configuration of this embodiment and the flow of information.

The structure of the city is represented by a graph structure, and the structure of the route and the value of the route are represented by an adjacency matrix (Ad (i, j)) (see FIGS. 2 and 3). The collected person trajectory data is stored in the behavior pattern table 1 and the profile table 2.

The word-of-mouth information exchange model 3 is a model that has information, and extracts features from city information by exchanging information and repeating evaluations. This model can be processed as follows, for example, by adding topographic conditions to the genetic algorithm.

The characteristic of the behavior pattern is that the path actually walked is expressed as a sequence of intersection numbers, and the behavior is inherited. That is, crossover and mutation are genetically manipulated for the two behavior patterns, and the generation change of the genes is advanced.
The process flow corresponding to the word-of-mouth information exchange model is as follows (see FIG. 6).

Step 1: First-generation behavioral genes are arranged, and the number of generations T is set to one.

Step 2: From the behavioral gene of the Tth generation, 2
Create individual sets and exchange information.

Step 3: In the Tth generation information exchange, two behavior genes of the T + 1 generation are generated by genetic crossover of the behavior pattern. When capturing information of a partner in genetic crossover, geographical constraints are performed with a certain probability. The evaluation value (Ad (i, j)) of a path (behavior pattern) that does not change as a result of genetic crossover is increased.

Step 4: Under certain conditions, some of the T + 1 generation behavior genes are mutated.

Step 5: The number of generations T is incremented by 1 and steps 2 and thereafter are repeated until the number of generations to be terminated Te is reached.

By this repetitive processing, excellent behavior genes survive natural selection, and the evaluation value (Ad (i, j)) of the path of the behavior pattern increases. When the number of generations to be finished Te is reached, the data is totaled and analyzed for each route and intersection. Since the result is, so to speak, a simulation of the flow of people in the entire city, the flow of people is directly displayed on a map instantaneously from time to time on the analysis terminal. In places where there are many people, a visual representation of how many people actually move on the screen is displayed (see FIG. 12). Also, for specific routes and intersections,
The characteristics of the passersby are analyzed.

FIG. 2 is a simple example of a town structure.

When determining a route, it is often the case that a map is graphed, intersections on the map are treated as nodes, and roads between the intersections are treated as edges. That is, a map is often treated as a graph represented by a set of nodes and a set of edges. Here, the graph structure of the city represented by the intersection P and the road E is represented by using the adjacency matrix Ad.

FIG. 3 is a diagram showing an adjacency matrix corresponding to the city in FIG.

The initial value of Ad (i, j) is 1 if there is a route from the i-th intersection P (i) to the j-th intersection P (j), and 0 if there is no route. A (i, i) = 1. Note that this adjacency matrix is sequentially rewritten with a value indicating the value of the route in the subsequent processing.

FIG. 4 is a diagram for explaining an action pattern table for storing collected trajectory data of a person.

The behavior pattern table stores behavior genes, and has a field composed of an ID, a behavior pattern, and a matching value. The ID is a number uniquely assigned to a record, and is used to correlate with the data in the profile table in FIG. The action pattern is expressed as a variable-length vector in the sense of the arrangement of the intersections P (i). In the order of the walking path, for example, $ P
(1), P (3), P (7), P (10), P (13), P
(15) When passing through the intersection in the order of｝, {1, 3, 7,
It is described as 10, 13, 15 °. It may be described together with the transit time. The adaptation value is a numerical value of an evaluation index that becomes a large number if the behavior pattern is stable.

For example, in the record of ID = 3, assuming that a department store (P (8)) at a certain place performs a bargain sale on a certain day, a person heading for the department store from the nearest station (P (5)) In many cases, they go straight and there are few elements that come to a different place on a whim. In the present embodiment, a behavior gene having a high evaluation index can be added to collected human trajectory data, and a simulation assuming such a special environmental change can be performed. In FIG. 4, the case of actually collected data is set to 1.

FIG. 5 is a profile table for storing attribute data of a person having a behavior gene.

It has fields composed of ID, age, gender, and address. The ID is a number uniquely assigned to the record, and is used to correlate with the data of the action j pattern table in FIG. It is used to analyze the flow of people by age and gender, and to analyze the flow of local people and the flow of people from distant visitors.

FIG. 6 shows the main processing for analyzing the flow of people using the word-of-mouth information exchange model.

In step 601, 2m first-generation behavioral genes are arranged, and the number of generations T is set to one.

In step 602, two sets are created from the Tth generation behavior gene, and information is exchanged.

In step 603, two (T + 1) th generation behavioral genes are generated by genetic crossover of behavioral patterns in the Tth generation (see FIG. 7).

In step 604, a mutation is caused in the behavior gene of the (T + 1) th generation under certain conditions that are spatially restricted with a certain probability. In other words, care is taken so that the route does not suddenly jump to another point that cannot be reached (see FIG. 10).

At step 605, it is determined whether or not the processing for the m sets of behavioral genes has been completed. If not, the procedure returns to step 603 to repeat the processing.

If so, in step 606,
The number of generations T is incremented by one, and what was the T + 1 generation is treated as the T generation again.

In step 607, the value of the route evaluation value Ad (i, j) is increased for the road whose route has not changed, as compared to the behavior gene whose generation has been changed before the change. This is a mechanism in which the path that converges as a result of the repetitive processing by the genetic algorithm is evaluated highly.

In step 608, the number of generations Te that is set to end in advance and T are compared. If T is less than Te, the process returns to step 602 to repeat the processing.
Otherwise, the repetition processing is terminated, and based on the obtained result, in step 609, the analysis and the result output are performed for each route intersection (see FIG. 11).

FIG. 7 is a diagram for explaining a process of performing a genetic crossover of a behavior gene.

In step 701, two sets of A and B are read as a set of Tth generation behavior genes. In step 702, the action patterns of A and B are sequentially compared,
It is checked whether there is an intersection (common intersection) through which both pass. If there is a common intersection, it can be interpreted to mean that there are other people who are performing close actions. Therefore, in step 703, the matching value of the behavior gene is increased. Further, in step 704, information is mutually exchanged, and a process of exchanging the destination after passing through the common intersection is performed (see FIG. 8 destination exchange).

In step 702, if there is no common intersection, in step 705, a point where the routes are closest to each other is selected, and a detour information exchange process for exchanging information so as to depart from each other is performed (FIG. 9). Information exchange). These two types of genetic manipulations are based on the word-of-mouth information exchange model / “Exchanging information on paths that have been walked with each other. Steps that may be interesting are taken in, and then used when visiting the city next time. It shall be referred to as "." After the above gene crossover process,
The newly generated offspring of A and B are recorded as T + 1 generation behavioral genes.

FIG. 8 is a diagram for explaining the processing contents of the destination exchange in the gene crossover processing using an example.

In FIG. 8, the behavior pattern of the behavior gene A of the Tth generation is represented by A (T), and the path is illustrated. Similarly, the behavior pattern of the behavior gene B of the Tth generation was set to B (T). By the destination exchange processing, attention is paid to P (7), which is a common intersection between the two, and subsequent destinations are exchanged between A and B. Based on the exchanged result, a behavior gene of the (T + 1) th generation is generated as behavior patterns A (T + 1) and B (T + 1).

FIG. 9 is a diagram for explaining, by way of example, the details of the process of exchanging detour information in the gene crossover process.

In FIG. 9, similarly to FIG. 8, the behavior pattern of the behavior gene A of the Tth generation is represented by A (T), and the behavior pattern of the behavior gene B is represented by B (T). By exchanging detour information, the two action patterns are compared, and a set of intersections ((p1a, p1b), (p2
a, p2b)), are selected. This processing can be performed in a sufficiently real time even if the whole case search is used (on the order of the square of the length of the action pattern). Then, the routes of the other party are taken in at these intersections so that the routes are exchanged. At this time, the vehicle passes through a new route that both have not passed. That is, the route is between the route (p1a-p1b) and the route (p2a-p2b).

The route search between two specific points in the graph structure data expressed by using an adjacency matrix including a loop as in the present embodiment is a known technique, and has been put to practical use in, for example, a car navigation system. I have. Since the result of the detour information exchange is a way of going through a road unknown to both sides, it is not possible to say that the road is good, so that the value of the adaptation value is not changed here.

Based on the result of the exchange, the behavior pattern A
The generation of the (T + 1) th generation behavior gene as (T + 1) and B (T + 1) is the same as in FIG.

FIG. 10 is a flowchart for explaining the mutation process.

In step 1001, a behavior gene is read. In step 1002, it is determined whether this gene is mutated. In addition to the predetermined occurrence frequency α, the probability is adjusted so that the smaller the value of the adaptation value is, the easier the mutation is to occur. When the mutation occurs, in step 1003, an appropriate random number is generated, and two points are randomly extracted from the behavior pattern in the behavior gene.

In step 1004, the shortest path between these two points is calculated. In the present embodiment, it is not an optimization problem that narrows down a person's behavior pattern to one.
A genetic algorithm is used to perform the trend analysis, and in this case, the solution is not necessarily shortest but may be a sub-optimal solution. In step 1005, the route data between the two points is replaced with the calculated shortest route. In step 1006, the behavior gene after mutation is recorded.

Note that this genetic manipulation is intended not to create an unusual gene, but rather to work on returning the unusual route information that can be generated as a result of the detour information exchange shown in FIG. 9 to a common sense route. Things.

FIG. 11 is a flowchart of a process for analyzing and outputting the results for each route and intersection. FIG.
Is an example of an output result screen, and the contents of the processing will be described below together.

In step 1101, the adjacency matrix Ad
(I, j) is read. In step 1102, the extreme value in the Z-axis direction is calculated by regarding the latitude and longitude direction of the ground plane as the XY axes and the value of the adjacent matrix as the Z axis. This XYZ three-dimensional space indicates the evaluation value of the route.
As shown in FIG. 2A, the evaluation value of a route with many traffic is high. Therefore, the extreme value in the Z-axis direction can be said to be a central road in the neighborhood.

In step 1103, if some thresholds are determined based on the obtained extreme values, for example,
As shown in FIG. 12B, a road showing an extreme value can be painted thick. Here, the reason why the extreme value is calculated instead of the maximum value is that, for example, a route with little traffic but a valuable route, such as a back street, can be detected with the extreme value. Of course, the maximum value is also important, but usually the busiest road can be calculated from simple statistics from fixed point observations.

In step 1104, the pedestrians of the route showing the extreme value are analyzed by referring to the profile table. In step 1105, the result is output. For example, "People passing through Station 1 to Station 2 are 2
0.9% "," 60.6% of people in a residential area pass (or have passed) on this road ". Further, in step 1106, the operation path of the behavior gene of the last generation is output. This result may be output, for example, as shown in FIG. 12 (c), in which individual behavior genes are directly operated on a map.

[0072]

According to the present invention, trajectory information obtained by continuously acquiring position information of a person is collected and analyzed to output the tendency of the flow of the person in more detail. Can be. Here, the flow of people is information on the number of people passing through a certain route, time distribution, and the like. As a result of the analysis, for example, “people who pass station 1 to station 2
%. And "60.6% of people in a residential area pass (or have passed) on this road."

Further, according to the word-of-mouth information exchange model, even if the collected trajectory information is that of a part of a person walking in the city,
The whole flow can be inferred. Visualize the flow of people and enable more detailed analysis than before.

By mixing artificial data indicating environmental changes, instead of actually acquired trajectories of people, in part, it is possible to simulate the flow of people throughout the city. For example, when a new road is passed, the route evaluation can be performed by changing the value of the corresponding portion of the adjacent matrix representing the city structure from 0 to 1. That is, it is possible to provide a flow analysis method of people that can flexibly respond to environmental changes.

As a result of the analysis, the flow of people can be directly displayed on a map moment by moment, and in places where there are many flows of people, it is possible to visually represent how many people actually move on the screen. Also, for specific routes and intersections,
The characteristics of the passersby can be analyzed.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an overall configuration and a flow of information according to an embodiment.

FIG. 2 is a diagram showing a simple example of a town structure.

FIG. 3 is a diagram showing an adjacency matrix corresponding to the town in FIG. 2;

FIG. 4 is a diagram illustrating an action pattern table that stores collected human trajectory data.

FIG. 5 is a diagram showing a profile table for storing attribute data of a person having a behavior gene.

FIG. 6 is a flowchart illustrating a flow analysis of people using a word-of-mouth information exchange model.

FIG. 7 is a flowchart illustrating a process of performing genetic crossover of a behavior gene.

FIG. 8 is a diagram illustrating the contents of destination exchange processing in gene crossover processing using an example.

FIG. 9 is a diagram illustrating, using an example, processing contents of detour information exchange in gene crossover processing.

FIG. 10 is a flowchart illustrating a mutation process.

FIG. 11 is a flowchart of a process for performing analysis and outputting results for each route and intersection.

FIG. 12 is a diagram illustrating an example of an output result screen.

[Explanation of symbols]

1: Behavior pattern table, 2: Profile table, 3: Word-of-mouth information exchange model.

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Shuji Kitazawa 4-6 Kanda Surugadai, Chiyoda-ku, Tokyo F-term in Hitachi Systems Co., Ltd. F-term (reference) 2F029 AA07 AB07 AC02 AC06 AC20 5H180 AA21 BB05 BB15 DD02 FF04 FF10 FF27 9A001 FF08 GG05 KK37 KZ60

Claims (2)

[Claims] In a computer system provided with a display, a memory, and a central processing unit, an intersection and a road of interest are selected for an area for investigating a flow of a person, and a graph is created in which the intersection is regarded as a node and the road is regarded as a branch. First step, second collecting monitor profile and trajectory data
Step, the third step of associating the latitude and longitude data of the trajectory collected by the monitor with the nodes and branches of the graph, the fourth step of aggregating data distribution trends based on a word-of-mouth information exchange model, and analysis by statistical processing And a fifth step of outputting to the display. 2. The method according to claim 1, further comprising a flow line analysis method using the collected trajectory data as a behavior gene, wherein the flow line analysis method includes arranging a first-generation behavior gene. In a first step in which the number of generations T is 1, a set of two sets of behavior genes of the Tth generation and a second step of exchanging information, in the information exchange of the Tth generation, A third step of generating two behavioral genes of the T + 1 generation by genetic crossing, a fourth step of performing geographical constraint when capturing information of the other party in the genetic crossing, and a result obtained from the genetic crossing. The fifth step of increasing the evaluation value of the pathway for the unchanged pathway in the behavior gene of the T + 1 generation, the T +
A sixth step of mutating some of the behavioral genes of one generation, a seventh step of incrementing the number of generations T by 1 and repeating steps 2 and thereafter until the number of generations to be terminated Te is reached Characteristic flow analysis method.