JP5930551B2 - Residence point extraction method, residence point extraction device, and residence point extraction program - Google Patents

Description

  The present invention relates to a technique for obtaining a user's stay point using position information of a user terminal.

  Location information acquired from a feature phone, a smartphone, or the like using GPS, a wireless base station, Wi-Fi, or the like is collected as positioning data, and the geographical behavior of the user is analyzed. And in order to understand where the user visited from the collected positioning data, there is a technique for extracting a point where the user stayed for a long time as a staying point using the positioning data. Specifically, the Mean-Shift technology, which is a kernel density-based clustering technique, extracts positioning points included in the positioning data, that is, locations where the points indicated by the longitude and latitude at the positioning time are dense as dwell points. (Non-Patent Document 1) is used.

Dorin Comaniciu, 1 other person, “Mean Shift: A Robust Approach Toward Feature Space Analysis”, IEEE TRANSACTION ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, Vol.24, No.5, May 2002, p.603-619

  However, since it is assumed that the positioning points are concentrated at the staying point, if the positioning acquisition interval changes due to, for example, suppressing the number of positionings when there is no movement from the viewpoint of reducing battery consumption such as feature phones, the staying point There is a problem that the number of positioning points near the point is reduced. In addition, positioning by GPS or the like varies depending on environmental factors such as indoors, outdoors, and radio wave conditions. Therefore, in the prior art that treats all the positioning points equally, there is a problem that the staying points are shifted by being drawn to the positioning points having a large measurement error. Due to these two problems, there is a possibility that the extraction accuracy of the staying point is lowered.

  This invention is made | formed in view of the said situation, and it aims at improving the extraction accuracy of the staying point where the user stayed.

In order to solve the above-mentioned problem, a dwell point extraction method according to claim 1 is a first step of storing position information collected from a user terminal as positioning data in a storage means by a computer, and each positioning data from the storage means. A second step of calculating the positioning accuracy of each positioning data using the positioning error data included in each positioning data, and setting each positioning point in each positioning data as a cluster, and from each cluster A third step of acquiring each of a plurality of positioning points within a certain region and having a positioning time interval within a certain time, and a weighting that uses the positioning accuracy and positioning data of the plurality of positioning points as a weight per calculates the center of gravity, and a fourth step of the center point of said each cluster, the distance between the center points of each cluster constant The fifth step of merging clusters that are less than or equal to the separation and the positioning points of each cluster included in the merged cluster are arranged in the order of the positioning time, and the time interval between the positioning times of two adjacent positioning points is a certain time or more In this case, the present invention includes a sixth step of separating the merged cluster and a fifth step of extracting a central point of the merged cluster and the cluster separated after the merge as a user's residence point. And

  According to the present invention, the positioning accuracy of each positioning data is calculated using positioning error data included in each positioning data, each positioning point in each positioning data is set as a cluster, and any given cluster within a certain region and Obtains multiple positioning points whose positioning time intervals are included within a certain time, calculates the weighted centroid using the positioning accuracy and positioning data of the multiple positioning points as weights, and the weighted centroid Is the center point of any cluster, and the center point is extracted as the user's stay point, so even if the positioning accuracy changes due to environmental factors, the location of the stay point will be shifted due to the influence of the positioning point with poor accuracy. Therefore, the staying point where the user stayed can be extracted with high accuracy.

  The stay point extracting method according to claim 2 is the stay point extracting method according to claim 1, wherein an interval between positioning times is larger than a predetermined time between the second step and the third step, and The gist of the present invention is to further include a step of adding new positioning data between positioning data whose positioning points are smaller than a certain distance.

  According to the present invention, since the new positioning data is added between the positioning data whose positioning time interval is larger than a certain time and between the positioning points is smaller than the certain distance, even when the positioning acquisition interval of the positioning data changes. Since a lot of positioning data can be interpolated at a point where the possibility of staying is high, the staying point where the user stayed can be extracted more accurately.

The stay point extraction apparatus according to claim 3 stores a storage unit that stores position information collected from a user terminal as positioning data, reads out each positioning data from the storage unit, and outputs positioning error data included in each positioning data. A plurality of calculation means for calculating the positioning accuracy of each positioning data, and each positioning point in each positioning data as a cluster, and within a certain area from each cluster and a positioning time interval within a certain time Obtaining means for obtaining each positioning point; and calculating means for calculating a weighted centroid weighted by positioning accuracy using the positioning accuracy and positioning data of the plurality of positioning points, and calculating the center point of each cluster; Merging means for merging clusters whose distance between the center points of each cluster is a certain distance or less, and each cluster included in the merged cluster. If the positioning points of the data are arranged in the order of the positioning times and the time interval between the positioning times of two adjacent positioning points is a certain time or more, the separating means for separating the merged cluster, the merged cluster, and after the merge The gist of the present invention is to include an extraction unit that extracts a center point of the separated cluster as a staying point of the user.

  The stay point extracting apparatus according to claim 4 is the stay point extracting apparatus according to claim 3, wherein new positioning data is obtained between positioning data in which the interval between positioning times is larger than a certain time and the distance between positioning points is smaller than a certain distance. The gist of the present invention is to further include an interpolation means for adding.

  A summary of the stay point extraction program according to claim 5 is to cause a computer to execute the stay point extraction method according to claim 1 or 2.

  ADVANTAGE OF THE INVENTION According to this invention, the extraction precision of the stay point where the user stayed can be improved.

It is a figure which shows the functional block structure of a stay point extraction apparatus. It is a figure which shows the structural example of positioning data DB. It is a figure which shows the processing flow of a positioning data extraction part. It is a figure which shows the processing flow of a positioning data interpolation part. It is a figure which shows the linear interpolation example of positioning data. It is a figure which shows the processing flow of a stay point extraction part. It is a figure which shows the structural example of residence point DB. It is a figure which shows the example of extraction of a stay point.

  Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing a functional block configuration of a stay point extraction apparatus 1 according to the present embodiment. The stay point extraction device 1 includes a positioning data DB 11, a positioning data extraction unit 12, a positioning data interpolation unit 13, a stay point extraction unit 14, and a stay point DB 15. Hereinafter, the function of each part will be described in detail.

  The positioning data DB 11 is a functional unit that collects and stores, as positioning data, position information that can be acquired from a user terminal such as a feature phone or a smartphone using a GPS, a portable radio base station, Wi-Fi, or the like.

  A configuration example of the positioning data DB 11 is shown in FIG. The positioning data includes a user ID (uid), positioning time (time), longitude (longitude), latitude (latitude), positioning method (method), positioning error (accuracy), and the like. The positioning error is an inaccuracy of positioning measured by GPS or the like, and is a parameter value already included in the collected position information. In the configuration example of FIG. 2, the unit is meters. Note that other information such as altitude and acceleration may be included in the positioning data.

  The positioning data extraction unit 12 extracts positioning data corresponding to the user ID described in the input target user ID list from the positioning data DB 11. At this time, the positioning accuracy is calculated from the positioning error according to the procedure of FIG. 3 described later, and is output after being added to the extracted positioning data.

  The positioning data interpolating unit 13 receives the positioning data output from the positioning data extracting unit 12, and in accordance with the procedure of FIG. 4 to be described later, the positioning time interval is larger than a certain time and the distance between positioning points is more than a certain distance. New positioning data is linearly interpolated between small positioning data, added to the original positioning data, and then output. The positioning point is a point indicated by longitude / latitude at the positioning time.

  The stay point extraction unit 14 receives the positioning data output from the positioning data interpolation unit 13, and performs a kernel density-based clustering method in which the positioning points are weighted according to the positioning accuracy according to the procedure of FIG. Use to extract the residence point. The extracted stay point is output to the stay point DB 15 and stored.

  Next, each operation | movement of the positioning data extraction part 12, the positioning data interpolation part 13, and the stay point extraction part 14 is explained in full detail. First, the positioning data extraction unit 12 will be described. FIG. 3 is a diagram illustrating a processing flow of the positioning data extraction unit 12.

  First, in step S101, the user ID included in the target user ID list given in advance is input, and all the positioning data corresponding to the unprocessed user ID is acquired from the positioning data DB 11. The subsequent step S102 is executed for each user ID.

  Next, in step S102, the positioning accuracy A (a) of each positioning data is calculated from equation (1) using each positioning error a included in each positioning data acquired in step S101.

ここで、Ａは測位精度を求める関数、ａは測位誤差の値、Ａ_０は測位誤差がゼロの場合の測位精度、λは減衰定数である。式（１）を用いることにより、測位精度Ａは、Ａ_０をピークとして測位誤差が大きくなるごとに指数関数的に減衰することになる。減衰の強さは減衰定数λの大きさによって調節できる。測位精度Ａ_０と減衰定数λは事前に人手によって任意に定めることができるが、共に正の値でなければならない。例えば、Ａ_０＝１、λ＝０．０１などを用いる。

Here, A is a function for obtaining positioning accuracy, a is a positioning error value, A ₀ is positioning accuracy when the positioning error is zero, and λ is an attenuation constant. By using the equation (1), positioning accuracy A will attenuate exponentially each time the positioning error becomes large as the peak A _0. The strength of the attenuation can be adjusted by the magnitude of the attenuation constant λ. The positioning accuracy A ₀ and the attenuation constant λ can be arbitrarily determined in advance by hand, but both must be positive values. For example, A ₀ = 1, λ = 0.01, etc. are used.

Note that the function is such that the positioning accuracy decreases as the positioning error increases, and a function other than Expression (1) may be used as long as it does not take a negative value. Also, a function that takes into account information other than the positioning error a, such as changing the positioning accuracy _A0 or the value of the attenuation constant λ according to a positioning method such as GPS, a wireless base station, or Wi-Fi, may be used.

  Then, the positioning accuracy calculated in this way is added to the positioning data acquired in step S101, and is output to the positioning data interpolation unit 13 as new positioning data.

  Thereafter, in step S103, it is determined whether or not there is an unprocessed user ID. If there is an unprocessed user ID, the process returns to step S101, and steps S101 and S102 are performed for all user IDs. On the other hand, if there is no unprocessed user ID, the process ends.

  Next, the operation of the positioning data interpolation unit 13 will be described. FIG. 4 is a diagram illustrating a processing flow of the positioning data interpolation unit 13.

  First, in step S201, the positioning data output from the positioning data extraction unit 12 is input, and all positioning data corresponding to an unprocessed user ID is acquired. The subsequent steps S202 to S205 are executed for each user ID.

Next, in step S202, unprocessed adjacent positioning data is acquired. Specifically, a positioning point and a positioning point that is temporally adjacent to the positioning point are acquired as adjacent positioning data from each positioning data acquired in step S201. Here, it is assumed that the set D of positioning data d is configured in ascending order of positioning time as {d ₀ , d ₁ ,..., D _n−1 | n> 2, n∈N}. A set D ′ including the above is expressed as {(d ₀ , d ₁ ), (d ₁ , d ₂ ),..., (D _n−2 , d _n−1 ) | n> 2, n∈N}. Note that n is the total number of positioning data corresponding to the user ID currently being processed. Thereafter, the adjacent positioning data is used to perform processing in order of ascending positioning time.

Next, in step S203, the adjacent positioning data d _i , d _{i + 1} (0 ≦ i <n−1; _i∈ ) using the positioning time interval threshold value T _lerr and the positioning distance interval threshold value L _lerrp determined in advance by hand. It is determined whether or not the positioning times t _i and t _{i + 1} and the positioning points l _i and l _{i + 1} included in Z) satisfy the condition of equation (2).

なお、ｄｉｓｔとは、与えられた２つの測位点ｌ_ｉ，ｌ_ｉ＋１間における、回転楕円体上の最小距離をメートル単位で出力する関数である。例えば、ヒュベニの距離計算式などを用いる。

The dist is a function that outputs the minimum distance on the spheroid between two given positioning points l _i and l _{i + 1} in meters. For example, the Huveni distance calculation formula is used.

If the condition of the expression (2) is satisfied, it is considered that the measurement number suppression process is working as described above. Therefore, the process proceeds to step S204, and between the two positioning data in the adjacent positioning data d _i and d _{i + 1} . Add new positioning data to. On the other hand, if the condition of formula (2) is not satisfied, the process proceeds to step S205 without performing step S204.

Next, in step S204, new positioning data is linearly interpolated at intervals of the interpolation interval threshold value S _lerrp . Specifically, when linear interpolation is performed between the adjacent positioning data d _i and d _{i + 1} , the interpolation positioning data D _{i, i + 1} calculated from the equation (3) is added as new positioning data.

すなわち、隣接する２つの測位点間の線分を補間間隔閾値Ｓ_ｌｅｒｐで分割し、分割された各点に補間測位点を新たに追加する。これにより、（Ｓ_ｌｅｒｐ−１）個の補間測位点が増加することになる。なお、ｌｏｎ_ｉは測位点ｌ_ｉにおける経度を表し、ｌａｔ_ｉは測位点ｌ_ｉにおける緯度を表している。図５は、測位データの線形補間例を示す図である。測位時間間隔閾値Ｔ_ｌｅｒｐ＝１５、測位距離間隔閾値Ｌ_ｌｅｒｐ＝２００、補間間隔閾値Ｓ_ｌｅｒｐ＝４の場合を例示している。

That is, a line segment between two adjacent positioning points is divided by the interpolation interval threshold value S _{l erp} , and an interpolation positioning point is newly added to each of the divided points. As a result, (S _{l erp} −1) interpolation positioning points are increased. In addition, lon _i represents the longitude at the positioning point l _i , and lat _i represents the latitude at the positioning point l _i . FIG. 5 is a diagram illustrating an example of linear interpolation of positioning data. The case where the positioning time interval threshold value T _{l erp} = 15, the positioning distance interval threshold value L l _erp = 200, and the interpolation interval threshold value S _{l erp} = 4 is illustrated.

  Next, in step S205, it is determined whether there is unprocessed adjacent positioning data. If there is unprocessed adjacent positioning data, the process returns to step S202, and steps S202 to S202 are performed for all unprocessed adjacent positioning data. S204 is performed. On the other hand, if there is no unprocessed adjacent positioning data, the process proceeds to step S206.

  Next, in step S206, it is determined whether or not there is an unprocessed user ID. If there is an unprocessed user ID, the process returns to step S201, and steps S201 to S205 are performed for all user IDs. On the other hand, if there is no unprocessed user ID, the process proceeds to step S207.

  Finally, in step S207, the positioning data that has been subjected to the linear interpolation by the processing so far, or the positioning data that has not been subjected to the linear interpolation is output to the stay point extraction unit 14.

  Next, the operation of the stay point extraction unit 14 will be described. FIG. 6 is a diagram illustrating a processing flow of the stay point extraction unit 14.

  First, in step S301, the positioning data output from the positioning data interpolation unit 13 is input, and all positioning data corresponding to an unprocessed user ID is processed. The subsequent steps S302 to S311 are executed for each user ID.

Next, in step S302, all the positioning points l included in the positioning data d acquired in step S301 are initialized as clusters, and the number of loops is set to zero. Here, the initialization means that the center point d _i ^c of each cluster is set to the value of the positioning point l _i itself. That is, the initial value of the center point d _i ^c of the i-th cluster is the positioning data d _i .

  Note that a cluster is a minimum unit of a range in which the user is considered to have stayed at a predetermined position. In Steps S302 to S308, each positioning point is set as a cluster, and then in Steps S309 and S310, the clustering of the clusters or the cluster after the merging is performed to extract the staying points of the users who remain at regular intervals.

  Next, in step S303, a plurality of pieces included within a radius R (unit: meters) and time T (unit: seconds) determined in advance by hand from the center of an unprocessed cluster (hereinafter, target cluster). Get a positioning point. Specifically, a set K of positioning point indexes included within a radius R and within a time T is calculated from Equation (4).

ここで、ｎは、現在処理中のユーザＩＤに該当する測位データの総数である。

Here, n is the total number of positioning data corresponding to the user ID currently being processed.

Next, in step S304, by using the positioning accuracy _{A (a k)} and positioning data _{d k} of the plurality of positioning points _{l k} obtained in step S303, the weighted centroid d _'i ^c to weight the positioning accuracy formula Calculate from (5).

ここで、重み付き重心ｄ’_ｉ ^ｃの算出方法を具体的に説明する。例えば、ステップＳ３０３で３つの測位データｄ_ｋが取得され、取得された各測位データｄ_１〜ｄ_３の実データが以下であるとする。右辺に示すデータ構造は、左から順に、測位時間，経度，緯度，測位精度である。

Here, concretely describing a method of calculating the weighted centroid d _'i ^c. For example, it is assumed that three positioning data d _k are acquired in step S303, and the actual data of the acquired positioning data d _{1 to} d ₃ is as follows. The data structure shown on the right side is, in order from the left, positioning time, longitude, latitude, and positioning accuracy.

d ₁ = {2014/1/1 00:00:00, [140, 40], 0.5}
d ₂ = {2014/1/1 00:10, [141, 41], 0.3}
d ₃ = {2014/1/1 00:20, [142, 42], 0.2}
Using these positioning data d _{1 to} d ₃ , weighted centroids for positioning time, longitude, latitude, and positioning accuracy are calculated, respectively, and all the calculated centroids are set as weighted centroids for the target cluster. For example, if the positioning time is t, t = (0.5 × [2014/1/1 00: 0] + 0.3 × [2014/1/1 00:10] + 0.2 × [2014/1/1. 00:20]) ÷ (0.5 + 0.3 + 0.2) = [2014/1/1 00:07] is the weighted positioning time.

  Similarly, in the case of longitude lon and latitude lat, [lon, lat] = (0.5 × [140, 40] + 0.3 × [141, 41] + 0.2 × [142, 42]) / (0 .5 + 0.3 + 0.2) = [140.7, 40.7] is the weighted longitude and latitude.

  In this way, the weighted centroid of multidimensional information such as longitude, latitude, and time included in the positioning data is calculated by substituting and calculating the positioning accuracy and the positioning data into Expression (5). In addition, information other than longitude / latitude and time, such as altitude, may be included in the positioning data, and in that case, the center of gravity may be further added.

Next, in step S305, the center point d _i ^c of the target cluster is moved to the weighted center of gravity d ′ _i ^c calculated in step S304. Here, the movement means changing the center point d _i ^c of the target cluster to the weighted center of gravity d ′ _i ^c . As a result, the center of the target cluster moves to a position corresponding to the positioning accuracy.

  Next, in step S306, it is determined whether or not there is an unprocessed cluster. If there is an unprocessed cluster, the process returns to step S303, and steps S303 to S305 are performed for all clusters. On the other hand, if there is no unprocessed cluster, the process proceeds to step S307.

Next, in step S307, for all the clusters, the process of changing the center d _i ^c of each cluster to the weighted centroid d ′ _i ^c is performed once, so the loop count is incremented and all the clusters are again registered. Unprocessed.

  Next, in step S308, it is determined whether or not the current loop count has reached the number of loop limit threshold I set in advance by hand, and if the loop limit threshold I has not been reached, Returning to step S303, this time, steps S303 to S305 are performed based on the moved center point moved in step S305.

  In the processing so far, for example, when there are 100 positioning data, 100 clusters are initially generated in step S302, and the center point of each cluster is moved to a position corresponding to the positioning accuracy in steps S303 to S305. Therefore, 100 clusters are gathered in a dense location of positioning data. On the other hand, if the loop limit number threshold I has been reached, the process proceeds to step S309.

  Next, in step S309, the clusters are merged to leave clusters as many as the number of places where the positioning data is dense. Specifically, clusters having a distance between cluster center points equal to or less than the cluster center point distance threshold C set in advance by hand are merged. Note that the distance between the center points of the cluster is calculated using the function dist. Further, the center point of the merged new cluster is calculated from the weighted centroid based on the equation (5). At that time, the union set of positioning points included in the cluster to be merged is used as an element of the index set K.

  Next, in step S310, if the positioning data set in the cluster is larger than the time interval threshold S set in advance by hand, the corresponding cluster is divided. That is, when the positioning points in the cluster are confirmed in order in ascending order of time, if the time interval between the time h and the time h + 1 is larger than the time interval threshold S, a cluster composed of positioning data up to the time h, Separated into clusters composed of positioning data after time h + 1. At this time, the position of the center point of each of the two separated clusters is set to the position of the center point of the original cluster. By this processing, clusters that have been merged from the viewpoint of distance in step S309 and become the same place on the way to return can be distinguished from the viewpoint of time.

  Next, in step S311, the center point of each remaining cluster is output as the user's stay point and stored in the stay point DB 15. A configuration example of the stay point DB 15 is shown in FIG. The configuration example of FIG. 7 illustrates a case where a certain user stays at one point. The residence start time (begin_time) and residence end time (end_time) at the residence point are defined by the minimum time and the maximum time of the positioning points included in the cluster, respectively. The residence time (stay_time) is obtained by subtracting the residence start time (begin_time) from the residence end time (end_time). Longitude and latitude are the center points of the cluster, that is, the user's residence point.

  In step S312, it is determined whether there is an unprocessed user ID. If there is an unprocessed user ID, the process returns to step S301, and steps S301 to S311 are performed for all user IDs. On the other hand, if there is no unprocessed user ID, the process ends.

  As described above, according to the present embodiment, the positioning data extraction unit 12 calculates the positioning accuracy of each positioning data using the positioning error included in each positioning data, and the staying point extraction unit 14 calculates each positioning data. Each positioning point in the data is set as a cluster, a plurality of positioning points included within a radius R and within a time T are acquired from the target cluster, and the positioning accuracy and the positioning data are used as the positioning accuracy. The weighted center of gravity is calculated, the weighted center of gravity is used as the center point of the target cluster, and the center point is extracted as the user's staying point, so even if the positioning accuracy changes due to environmental factors, It can suppress that the place of a stay point shifts by influence.

  Further, according to the present embodiment, the positioning data interpolation unit 13 adds new positioning data between the positioning data whose positioning time interval is larger than the positioning time interval threshold and between the positioning points is smaller than the positioning distance interval threshold. Therefore, even when the positioning acquisition interval of the positioning data changes, a lot of positioning data can be interpolated at a point where the possibility of staying is high.

  By performing the two processes described above, the staying point where the user stayed can be extracted with high accuracy as shown in FIG.

  Finally, the stay point extraction device 1 described in the present embodiment can be realized by a computer having a memory and a CPU. It is also possible to construct each operation of the stay point extraction apparatus 1 as a program, install it in a computer and execute it, or distribute it via a communication network.

1 ... Residence point extraction device 11 ... Positioning data DB
DESCRIPTION OF SYMBOLS 12 ... Positioning data extraction part 13 ... Positioning data interpolation part 14 ... Residence point extraction part 15 ... Residence point DB
S101 to S103, S201 to S207, S301 to S312 ... step

Claims (5)

By computer
A first step of storing position information collected from a user terminal in a storage means as positioning data;
A second step of reading each positioning data from the storage means and calculating positioning accuracy of each positioning data using positioning error data included in each positioning data;
Wherein each positioning points in each measurement data and each cluster, a third step of spacing constant region and positioning time from each cluster to obtain a plurality of positioning points included within a predetermined time,
A fourth step of calculating a weighted centroid having a positioning accuracy as a weight using the positioning accuracy and positioning data of the plurality of positioning points, and setting it as a center point of each of the clusters;
A fifth step of merging clusters whose distance between the center points of each cluster is less than or equal to a certain distance;
A positioning point of each cluster included in the merged cluster is arranged in the order of the positioning time, and when the time interval between the positioning times of two adjacent positioning points is equal to or longer than a predetermined time, the sixth cluster is separated. Steps,
A seventh step of extracting the central point of the merged cluster and the cluster separated after the merge as a user residence point;
A method for extracting staying points, comprising:   A step of adding new positioning data between positioning data between the positioning time and the positioning time interval larger than a certain time and between the positioning points smaller than the certain distance between the second step and the third step. The dwell point extraction method according to claim 1, comprising: Storage means for storing position information collected from the user terminal as positioning data;
Calculation means for reading each positioning data from the storage means and calculating positioning accuracy of each positioning data using positioning error data included in each positioning data;
Each positioning point in each positioning data as a cluster, respectively , an acquisition means for respectively acquiring a plurality of positioning points that are included in a certain area and within a certain time interval between positioning times from each cluster;
Calculating a weighted centroid with a positioning accuracy as a weight using the positioning accuracy and positioning data of the plurality of positioning points, and calculating means as a center point of each of the clusters;
A merging means for merging clusters whose distance between the center points of each cluster is equal to or less than a certain distance;
Separating means that separates the merged clusters when the positioning points of the clusters included in the merged cluster are arranged in the order of the positioning time, and the time interval between the positioning times of two adjacent positioning points is equal to or greater than a certain time. ,
An extraction means for extracting the central point of the merged cluster and the cluster separated after the merge as a staying point of the user;
A dwell point extraction device characterized by comprising:   4. The stay point extraction according to claim 3, further comprising interpolation means for adding new positioning data between positioning data having a positioning time interval larger than a certain time and between positioning points smaller than a certain distance. apparatus.   A stay point extraction program that causes a computer to execute the stay point extraction method according to claim 1.

Images