JP6360000B2 - Moving means estimation model generation apparatus, moving means estimation model generation method, moving means estimation model generation program - Google Patents

Description

  The present invention relates to a technique for generating a moving means estimation model for estimating moving means based on a GPS trajectory.

  With the spread of smartphones, the use of user location information acquired by GPS positioning and the like, and the provision of information according to the user's situation and the life log service that records the user's behavior have come to be frequently used. It was. Accordingly, there is a demand for a technique for estimating the user situation from the user position information.

  One of related techniques is moving means determination. This moving means determination is performed by extracting a section (segment) that is the same moving means with respect to a series of measurement points (GPS trajectory) having position information and time information acquired by GPS positioning or WiFi positioning, and moving with respect to the segment. Estimation of means (for example, walking, automobile, train, etc.) is performed.

  Thus, the moving means determination is composed of two phases: (1) segment extraction and (2) moving means estimation. (1) As for segment extraction, there are a method of dividing a section at a fixed time, a method based on a change point using a speed and acceleration information obtained from a GPS trajectory, and the like. (2) For moving means estimation, a method is used to construct a moving means prediction model in a supervised learning framework by using moving means annotations for GPS trajectories (Non-patent Documents 1 and 2). .

  In moving means estimation, it is important how a feature quantity effective for prediction can be acquired from information consisting of simple positioning points in a segment. Non-Patent Document 1 and Non-Patent Document 2 extract the movement distance, speed, acceleration, speed change rate, stop rate, and direction change rate of the GPS trajectory in the segment as feature amounts. Furthermore, in Non-Patent Document 3, the feature amount is extracted from the image format data representing the movement trajectory using deep learning, and the feature amount is used for the prediction of the movement unit, thereby further improving the accuracy of the movement unit estimation. The technique to make it disclosed is disclosed.

Zheng, Y., Liu, L., Wang, L., and Xie, X .: Learning transportation mode from raw GPS data for geographic applications on the web, In Proc. Of WWW'08, pp. 247-256, 2008 . Zheng, Y., Chen, Y., Li, Q., Xie, X., and Ma, W.-Y .: Understanding transportation modes based on GPS data for web applications, ACM Trans. Web, Vol. 4, No .1, pp.1: 1-1: 36, 2010. Endo Yuuki, Yoshihiko Kazuhara, Hiroyuki Toda, Yoshimasa Koike: “Feature Extraction from GPS Trajectory Using Representation Learning for Moving Means”, WebDB Forum 2014, 2014-11-19. Vincent, P., Larochelle, H., Lajoie, I., Bengio, Y. and anzagol, P.-A .: Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion.In Proc. Of JMLR '10, 11: 3371-3408, Dec. 2010. Rumelhart, D., Hinton, G., and Williams, R. (1986a). Learning representations by back-propagating errors. Nature, 323, 533-536.

  However, in the technique described in Non-Patent Document 3, no map image information on the segment is used as a feature amount used for moving means estimation. That is, when the moving means cannot be distinguished only by the feature amount based on the movement trajectory, information in the map image (for example, route shape and color information, landscape information of surrounding buildings, etc.) cannot be used as the feature amount, There was a problem that the estimation accuracy of the moving means did not increase.

  The present invention has been made in view of the above circumstances, and it is an object of the present invention to make it possible to determine a moving unit that cannot be determined by a conventional speed, acceleration, feature amount of a trajectory image, or the like.

  Therefore, the present invention uses a feature amount extracted from a map image corresponding to the GPS trajectory when generating the moving means estimation model for estimating the moving means based on the GPS trajectory.

  An aspect of the device of the present invention is a device that generates a moving device estimation model for estimating a moving device based on a GPS trajectory, and the positioning points acquired by GPS positioning for each section that is the same moving device. A first database for storing enumeration information as GPS trajectory segments; a second database for storing labeled GPS trajectory segments with labels on the GPS trajectory segments; and positioning points for each GPS trajectory segment stored in both databases When extracting the trajectory image from the sequence information, the trajectory image extraction unit that images the center of gravity of the sequence information of the positioning points as the center of the image, and the map image database that stores the map image, A map image acquisition unit for acquiring a map image corresponding to a GPS locus segment; and the first database. A non-labeled trajectory map image is generated based on the trajectory image extracted from the GPS trajectory segment and the map image corresponding to the GPS trajectory segment, while the trajectory image extracted from the GPS trajectory segment with the label of the second database and the relevant label exist A trajectory map image generation unit that generates a labeled trajectory map image based on a map image corresponding to a GPS trajectory segment, a third database that stores the unlabeled trajectory map image, and the labeled trajectory map image as the label. The DNN feature value is extracted from the learning result of the fourth database stored in correspondence with the third database and the storage data of the third database and the fourth database, and the extracted DNN feature value is associated with the label. DNN feature values stored in the fifth database A basic feature amount extraction unit that extracts a basic feature amount from a GPS trajectory segment with a label in the second database and stores the extracted basic feature amount in a sixth database in association with the label; And a seventh database that stores the linked feature values in association with the labels, and a feature value linking unit that connects the DNN feature values and the basic feature values corresponding to each of the five databases and the sixth database. An estimation model generation unit that generates the estimation model based on

  According to an aspect of the method of the present invention, a first database that stores, as a GPS trajectory segment, a list of positioning points acquired by GPS positioning for each section that is the same moving means, and a label attached to the GPS trajectory segment A second database for storing labeled GPS trajectory segments, and a computer for generating a moving means estimation model for estimating moving means, the positioning of each GPS trajectory segment stored in both databases; When extracting the trajectory image from the point sequence information, a trajectory image extraction step for imaging the center of gravity of the sequence information of the positioning points as the center of the image, and a map image database for storing the map image are stored in both databases. A map image acquisition step of acquiring a map image corresponding to each GPS locus segment; A trajectory image extracted from the labeled GPS trajectory segment of the second database while generating an unlabeled trajectory map image based on the trajectory image extracted from the GPS trajectory segment of the first database and the map image corresponding to the GPS trajectory segment. A trajectory map image generating step for generating a labeled trajectory map image based on the map image corresponding to the labeled GPS trajectory segment, and a trajectory map image storing step for storing the unlabeled trajectory map image in a third database; From the learning result of the multilayer neural network based on the third database and the stored data of the fourth database, DNN from the labeled locus map image storing step of storing the labeled locus map image in correspondence with the label in the fourth database. Extract features A DNN feature extraction step for storing the extracted DNN feature quantity in the fifth database in correspondence with the label; a basic feature quantity is extracted from the labeled GPS trajectory segment of the second database; and the extracted basic feature quantity The basic feature quantity extracting step for storing the data in the sixth database in association with the label, and the DNN feature quantity and the basic feature quantity respectively corresponding to the fifth database and the sixth database are connected to form a connected feature quantity. A feature amount connecting step; and an estimated model generating step for generating the estimated model based on a seventh database storing the connected feature amount in association with the label.

  Note that the present invention may be in the form of a program that causes a computer to function as each unit of the apparatus or a program that causes a computer to execute each step of the method.

  According to the present invention, it is possible to discriminate moving means that could not be discriminated based on the conventional speed, acceleration, feature amount of a trajectory image, and the like.

BRIEF DESCRIPTION OF THE DRAWINGS The whole block diagram of the moving means estimation model production | generation apparatus in embodiment of this invention. (A) is the figure which showed the visualization example of a map image, (b) is the data structure figure of the said example. The data structure figure of GPS locus segment DB without a label. The data structure figure of GPS locus segment DB with a label. (A) is the figure which showed the visualization example of a locus | trajectory image, (b) is the data structure figure of the said example. The flowchart which shows the process step of a map image acquisition part. The flowchart which shows the process step of a locus | trajectory map image generation part. The data structure figure of a locus map image. The data structure figure of a locus map image DB without a label. The data structure figure of locus map image DB with a label. The flowchart which shows the process step of a locus | trajectory image extraction part. The flowchart which shows the detail of S130 of FIG. The flowchart which shows the process step of a DNN learning part. The data structure figure of DNDB. The flowchart which shows the process step of a DNN feature-value extraction part. The data structure figure of DNN feature-value DB. The data structure figure of basic feature-value DB. The flowchart which shows the process step of a feature-value connection part.

  Embodiments of the present invention will be described below with reference to the drawings.

≪Overview≫
The moving means estimation model generation apparatus 1 shown in FIG. 1 uses the feature amount extracted from the map image corresponding to the GPS trajectory when generating the moving means estimation model for estimating the moving means based on the GPS trajectory. Thus, the moving means is determined with higher accuracy than in the past.

≪Example of device configuration≫
The moving means estimation model generation device 1 is configured by a computer and includes hardware resources such as a CPU, a main storage device (RAM, ROM, etc.), and an auxiliary storage device (hard disk drive device, solid state drive device, etc.).

  As a result of the cooperation between the hardware resource and the software resource (OS, application, etc.), the moving means estimation model generation device 1 has an unlabeled GPS trajectory segment DB10, a labeled GPS trajectory segment DB20, a trajectory image extraction unit 30, and a map image. DB 160, map image acquisition unit 170, trajectory map image generation unit 180, unlabeled trajectory map image DB 40, labeled trajectory map image DB 50, DNN learning unit 60, DNN DB 70, DNN feature amount extraction unit 80, DNN feature amount DB 90, basic features An amount extraction unit 100, a basic feature amount DB 110, a feature amount connection unit 120, a connection feature amount DB 130, an estimation model generation unit 140, and an estimation model DB 150 are mounted.

  The unlabeled GPS trajectory segment DB 10 stores, as GPS trajectory segments, enumeration information of positioning points acquired by GPS positioning for each section which is the same moving means.

  The labeled GPS trajectory segment DB 20 stores a labeled GPS trajectory segment with a label attached to the GPS trajectory segment.

  When the trajectory image extraction unit 30 extracts a trajectory image from the enumeration information of the positioning points of the GPS trajectory segments stored in the DBs 10 and 20, the trajectory image extraction unit 30 images the center of gravity of the enumeration information of the positioning points as the center of the image.

  The map image acquisition unit 170 acquires a map image corresponding to each GPS trajectory segment stored in the DBs 10 and 20 from the map image DB 160 storing the map image. Specifically, the longitude and latitude of the barycentric coordinates of the GPS trajectory are calculated based on the GPS trajectory acquired from the DBs 10 and 20 at regular time intervals, and the longitude width centered on the coordinates composed of the longitude and latitude, A map image included in the range of the latitude width is acquired from the map image DB 160.

  The trajectory map image generation unit 180 generates an unlabeled trajectory map image based on the trajectory image extracted from the GPS trajectory segment of the DB 10 and the map image corresponding to the GPS trajectory segment. Specifically, the unlabeled trajectory map image is configured by combining a vector indicating the pixel element of the trajectory image extracted from the GPS trajectory segment of the DB 10 and a vector indicating the pixel element of the map image corresponding to the segment. A vector indicating the element of the pixel to be generated is generated. The unlabeled locus map image is stored in the unlabeled locus map image DB 40.

  Further, the trajectory map image generation unit 180 generates a labeled trajectory map image based on the trajectory image extracted from the labeled GPS trajectory segment in the DB 20 and the map image corresponding to the labeled GPS trajectory segment. Specifically, the labeled trajectory map image is configured by combining a vector indicating a pixel element of a trajectory image extracted from the labeled GPS trajectory segment of the DB 20 and a vector indicating a map image pixel corresponding to the segment. A vector indicating the element of the pixel to be generated is generated. The labeled trajectory map image is stored in the labeled trajectory map image DB 50 in association with the label.

  The DNN feature quantity extraction unit 80 extracts the DNN feature quantity by extracting the learning result of the multilayer neural network (Deep Neural Network) by the DNN learning unit 60 based on the data stored in the DBs 40 and 50 from the DNN DB 70. The DNN feature value is stored in the DNN feature value DB 90 in association with the label.

  The basic feature amount extraction unit 100 extracts a basic feature amount from the labeled GPS trajectory segment of the DB 20, and stores the extracted basic feature amount in the basic feature amount DB 110 in association with the label.

  The feature amount linking unit 120 connects the DNN feature amount and the basic feature amount respectively corresponding to the DNN feature amount DB 90 and the basic feature amount DB 110 to obtain a connected feature amount. The connected feature value is stored in the connected feature value DB 130 in association with the label.

  The estimation model generation unit 140 generates a moving means estimation model for estimating moving means based on the connected feature values in the connected feature value DB 130. The moving means estimation model is stored in the estimation model DB 150.

  Each of the DBs 10, 20, 40, 50, 70, 90, 110, 130, 150, and 160 is constructed in the storage device. The moving means estimation model generation apparatus 1 may be configured as a single computer or a plurality of computers.

  Hereinafter, the details of each of the configurations 10 to 180 will be described.

≪Unlabeled GPS track segment DB10≫
The data structure of the DB 10 will be described with reference to FIG. The DB 10 stores list information of positioning points acquired by GPS positioning for each section (segment) which is the same moving means. The enumeration information of the positioning points is shown as a GPS locus segment in FIG.

  This GPS trajectory segment holds information in which latitude / longitude / positioning time is paired as enumeration information of the positioning points. The segment ID in FIG. 3 indicates the ID assigned to the GPS trajectory segment, and there are no particular restrictions on the unit of the ID.

≪Labeled GPS track segment DB20≫
The data structure of the DB 20 will be described with reference to FIG. In the DB 20, in addition to the GPS trajectory segment and the segment ID, label information of the moving means for each GPS segment is stored.

<< Trace image extraction unit 30 >>
The trajectory image extraction unit 30 reads data from each of the DBs 10 and 20 and extracts a trajectory image from each data. First, the overall processing steps of the trajectory image extraction unit 30 will be described with reference to FIG. Here, the trajectory image extraction unit 30 receives all GPS trajectory segments from the DB 10 when the process is started (S110). Subsequently, all label information and GPS trajectory segments are received from the DB 20 (S120).

Then the enumeration information of the positioning points in each GPS path segment s acquired from the DB10,20 and P _s, extracts the path image from P _s (S130). Then, the trajectory image I and the segment ID are transferred to the trajectory map image generation unit 180. If there is a label, the label is also delivered (S140).

  Next, the central processing content of the trajectory image extraction unit 30, that is, the trajectory image extraction step of S130 will be described in detail with reference to FIG.

S201: When the process starts the enumerated information P _s of the positioning point to produce a P _s' sampled at intervals of T seconds. Here, T is an arbitrary parameter.

S202: Calculate the longitude component center _lng of the barycentric coordinates in the enumeration information of the positioning points in the latitude / longitude coordinate system. Specifically, the calculation is performed according to the equation (1). | P _s '| in the equation (1) indicates the number of elements of P _s '. P (j). lng represents the longitude of the j-th positioning point at P _s ′.

S203: Calculate the latitude component center _lat of the barycentric coordinates in the positioning information in the latitude / longitude coordinate system. Specifically, the calculation is performed according to Expression (2). In this equation (2), p (j). “lat” represents the latitude of the j-th positioning point at P _s ′.

S204: The longitude coordinate min _{lng of the} coordinate at the most southwest in P _s ′ is calculated. Specifically, the calculation is performed according to Equation (3).

S205: The latitude coordinate min _{lat of the} coordinate at the most southwest in P _s ' is calculated. Specifically, the calculation is performed according to Equation (4).

S206: Calculate a correction value offset _x used to make the longitude coordinate of the barycentric coordinate in the arrangement information of the positioning points in the latitude / longitude coordinate system coincide with the horizontal component of the center coordinate of the image coordinate system. Specifically, the calculation is performed according to Equation (5).

Here, W _m is an arbitrary parameter representing the horizontal width of the trajectory image. Also, ε is a constant, and an arbitrarily small value such as 10 ⁻⁹ is used.

S207: latitude and longitude coordinates latitude coordinate of the barycentric coordinates in enumerated information of the positioning points in the system, to calculate a correction value offset _y used to match the vertical component of the center coordinates of the image coordinate system. Specifically, the calculation is performed according to Equation (6). H _m in this equation (6) is an arbitrary parameter representing the vertical width of the trajectory image.

S208: A W _m × H _m matrix I representing a trajectory image to be extracted is prepared, all elements are initialized with 0, and then the process proceeds to a loop process of S209 to S212. In this loop processing, values from 1 to | P _s ' | are sequentially substituted for the order j of positioning points. Hereinafter, the loop processing of S209 to S212 will be described.

S209: p ^(j) . Lng is converted from the longitude component of the latitude-longitude coordinate system to the horizontal component x of the image coordinate system. This conversion is performed by the calculation of Expression (7). W _p in this equation (7) indicates the longitude width when imaging with the center of gravity of P _s ′ as the center of the image.

S210: p ^(j) . lat is converted from the latitude component of the latitude-longitude coordinate system to the vertical component y of the image coordinate system. This conversion is performed by the calculation of Expression (8). In this equation (8), H _p indicates the latitude width when imaging with the center of gravity of P _s ′ as the center of the image.

  S211: If the coordinate x converted in S209 or the coordinate y converted in S210 is within the determined image size range, the process proceeds to S212, and if not, S212 is skipped.

  S212: A constant is added to the pixel where the positioning point exists. Here, the calculation of Expression (9) is performed.

I _{(x, y} ) in the equation (9) indicates an element of x rows and y columns of the matrix I. The pixel value is adjusted according to the staying time at the same point by the addition processing of Equation (9). For example, if the user stays at the same point for a long time, the corresponding pixel value of the trajectory image increases, and thus the staying time can be determined. If all the elements of | P _s ' | are substituted for j and the processing of S209 to S212 is completed, the loop processing is terminated.

≪Map image DB 160≫
The map image DB 160 stores map images. A map image, as shown in FIG. 2, is an arbitrary image with color information representing features (buildings and roads) at a certain scale. Images generated from data and the like are also included. The data structure is represented by a matrix as shown in FIG. 2, and each element of the matrix has a vector consisting of the number of dimensions for representing color information. For example, a three-dimensional RGB color space can be used as the color space for the map image.

≪Map image acquisition unit 170≫
The map image acquisition unit 170 reads data from the unlabeled GPS track segment DB 10 and the labeled GPS track segment DB 20. As shown in FIG. 3, the data structure of the unlabeled GPS trajectory segment DB 10 has a segment ID and GPS trajectory segment information. The data structure of the labeled GPS trajectory segment DB 20 has segment ID, label, and GPS trajectory segment information as shown in FIG.

  A map image corresponding to the coordinates of the read data is read from the map image DB 160 and enlarged or reduced to the same size as the trajectory image extracted by the trajectory image extraction unit 30. Here, in the trajectory image, a route that the user has passed is represented by a straight line or a curve as shown in FIG. The data structure is represented by a matrix as shown in FIG. 5, and each element of the matrix has a real value as a stay time in the pixel.

  FIG. 6 is a flowchart showing the flow of processing of the map image acquisition unit 170. The processing flow of the map image acquisition unit 170 will be described with reference to FIG.

S610: The segment ID and the GPS trajectory Ps in the segment _s are received from the unlabeled GPS trajectory segment DB10 or the labeled GPS trajectory segment DB20. If a label exists, it will also receive a label.

S620: In order to reduce the error in the positioning time interval of the GPS trajectory, the GPS trajectory P _s is sampled at intervals of T seconds to generate P _s ′. Here, T is a constant, and an arbitrary real value such as 60 is used.

S630: The longitude center _lng of the barycentric coordinates of the GPS trajectory P _s is calculated as in equation (1).

S640: The latitude center _lat of the barycentric coordinates of the GPS trajectory P _s is calculated as in equation (2).

_{S650: (center lng, center lat} ) and longitude width W _p centered on, the scale scale map image included in the scope of the latitudinal width H _p requests the map image DB 160, the map image I _m of the map image DB 160 is applicable Send. Here, W _p and H _p use the same value as the trajectory image extraction unit 30. The scale is set to an arbitrary value such as 1/10000 or 1/5000.

S660: acquired map image I _m the width W _m pixels, enlarged or reduced so that the vertical width H _m pixels. Here, W _m and H _m use the same values as the trajectory image extraction unit 30.

S670: The map image I _m and the segment ID are transferred to the trajectory map image generation unit 180. If there is a label, the label is also delivered.

≪Track map image generation unit 180≫
The trajectory map image generation unit 180 receives a trajectory image or the like from the trajectory image extraction unit 30 and receives a map image or the like from the map image acquisition unit 170. Next, a trajectory map image is generated by combining the received trajectory image and the map image, and the map trajectory image or the like is stored in the labeled trajectory map image DB 50 or the unlabeled trajectory map image DB 40.

  FIG. 7 is a flowchart showing a processing flow of the locus map image generation unit 180. With reference to FIG. 7, the process flow of the locus map image generation unit 180 will be described.

S710: The map image I _m and the segment ID are received from the map image acquisition unit 170. If you have a label, you will receive it.

  S720: The trajectory image I and the segment ID are received from the trajectory image extraction unit 30.

  S730: A trajectory image having a matching segment ID and a map image are combined. Specifically, the vectors corresponding to the respective pixels are combined. For example, if the vector representing the one-pixel element of the trajectory image is one-dimensional and the vector representing the one-pixel element of the map image is three-dimensional, the one-pixel element obtained by combining them is a four-dimensional vector. I ′ is generated. FIG. 8 shows an example of the data structure of the locus map image I ′ generated.

  S740: It is confirmed whether a label is received from the map image acquisition unit 170 in S710.

  S750: When it is determined Yes in S740, the segment ID, the label, and the generated trajectory image I 'are stored in the trajectory map image DB 50 with a label.

  S760: When it is determined No in S740, the segment ID and the generated trajectory image I 'are stored in the unlabeled trajectory map image DB 40.

≪Label map image DB40 without label≫
The unlabeled trajectory map image DB 40 is a DB that stores the trajectory map image generated by the trajectory map image generation unit 180 and the segment ID. FIG. 9 shows an example of the data structure of the unlabeled locus map image DB 40.

≪Label map image DB50 with label≫
The labeled trajectory map image DB 50 is a DB that stores the trajectory map image generated by the trajectory map image generation unit 180, the segment ID, and the label. FIG. 10 shows an example of the data structure of the labeled locus map image DB50.

<< DNN learning unit 60, DNNNDB70 >>
The DNN learning unit 60 reads the trajectory map image and the label information from the DBs 40 and 50 and inputs them, and performs learning of a multilayer neural network (DEEP Neural Network: DNN).

  The processing steps of the DNN learning unit 60 will be described with reference to FIG. That is, when the process is started, a trajectory map image is first received from the DB 40 (S310). Next, label information and a trajectory map image are received from the DB 50 (S320).

  Subsequently, DN-Pre-training is performed using all the trajectory map images acquired in S310 and S320 (S330). As a pre-training method, a known technique such as “Stacked Denoising Autoencoder” of Non-Patent Document 4 can be used.

As a result of the pre-training, N weight matrices “W = (W ₁ ^T , W ₂ ^T ,..., W _m ^T ) ^T ” and a bias term b are output for a DNN with N intermediate layers Is obtained.

  Then, fine-tuning for adjusting the weight matrix W and the bias term b of the entire DNN is performed using the trajectory map image and label information acquired in S320 (S340). As the fine-tuning method, a known technique based on the error back propagation method of Non-Patent Document 5 or the like can be used.

  Thereafter, the information of the weight matrix W and the bias term b is stored in the DB 70 (S350), and the process ends. As a result, as shown in FIG. 14, the DB 70 stores a weight matrix W and a bias term b for each intermediate layer number as a learning result of the DNN learning unit 60.

<< DNN feature quantity extraction unit 80, DNN feature quantity DB 90 >>
The DNN feature quantity extraction unit 80 reads out the stored data of the DB 70 as an input, and executes DNN feature quantity extraction. Referring to FIG. 15, when the process is started, the DNN feature amount extraction unit 80 first receives the label information and the trajectory map image from the DB 50 (S410), and then the weight matrix in each intermediate layer from the DB 70. W and the bias term b are received (S420).

  Next, the feature amount of each trajectory image received in S410 is calculated from the DNN using the weight matrix W and the bias term b received in S420 (S430). As a method for calculating the feature amount, a known method used in Non-Patent Document 4 or the like can be used. For example, the output of the deepest intermediate layer of DNN obtained by applying the trajectory image to the DNN input layer can be used as the feature amount of the trajectory image.

  Then, a pair of the feature amount (hereinafter referred to as DNN feature amount) of each trajectory image calculated in S430 and label information corresponding to the trajectory image is stored in the DB 90 (S440), and the processing is terminated. . As a result, as shown in FIG. 16, the DB 90 stores label information and DNN feature values for each segment ID of the GPS trajectory segment.

<< Basic Feature Extraction Unit 100, Basic Feature DB 110 >>
As shown in FIG. 1, the basic feature quantity extraction unit 100 reads the data in the DB 20 and extracts a feature quantity (hereinafter referred to as a basic feature quantity) from the read data. As the basic feature extraction method, for example, a known technique (feature engineering or the like) described in Non-Patent Documents 1 and 2 can be used.

  The basic feature amount extracted here is stored in the DB 110 in association with the label information. Thereby, as shown in FIG. 17, the DB 110 stores the label information and the basic feature amount for each segment ID of the GPS trajectory segment.

<< Feature Quantity Linking Unit 120, Linked Feature Quantity DB 130 >>
The feature amount linking unit 120 reads the data in the DBs 90 and 110 and connects the basic feature amount and the DNN feature amount related to the same segment ID between the respective data. The feature quantity connected here is referred to as a connected feature quantity.

  The processing steps of the feature amount linking unit 120 will be described based on FIG. That is, when the processing is started, the feature amount linking unit 120 first receives the segment ID, label information, and DNN feature amount from the DB 90 (S510), and then receives the segment ID, label information, and basic feature amount from the DB 110. Is received (S520).

  Next, the feature amount linking unit 120 specifies the basic feature amount and the DNN feature amount corresponding to the segment ID between the data received in S510 and S520, and combines the specified basic feature amount and the DNN feature amount into one vector. (S530), and the connected feature quantity is stored in the DB 130 (S540). As a result, the connection feature amount is stored in the DB 130 for each label information.

<< Estimated Model Generation Unit 140, Estimated Model DB 150 >>
The estimation model generation unit 140 reads data from the DB 130, and generates a moving means estimation model using the connected feature amount in the read data and the label information corresponding to the connected feature amount. For generating the estimation model, for example, known techniques such as logistic regression, SVM, and decision tree can be used (see Non-Patent Documents 1 and 2). The moving means estimation model generated here is stored in the DB 150.

<< Effects of this embodiment >>
According to the moving means estimation model generation device 1 described above, a feature amount is extracted from a map image by deep learning, and is used for moving means estimation together with a conventional feature amount, thereby moving more accurately than in the past. The means can be estimated. Specifically, the amount of features effective for the moving means such as the color and shape of the route in the map image, the difference in the surrounding landscape, etc. are quantified and extracted by deep learning, and using this, the conventional speed, Moving means that could not be discriminated by the acceleration or the feature amount of the trajectory image can be discriminated.

≪Programs≫
The present invention is not limited to the above-described embodiment, and can be implemented by being modified and applied within the scope described in each claim. For example, this invention can also be comprised as a moving means estimation model production | generation program which makes a computer function as some or all of each part 10-180 of the movement means estimation model production | generation apparatus 1. FIG. According to this program, it is possible to cause the computer to execute a part or all of S110 to S140, S201 to S212, S310 to S350, S410 to S440, S510 to S540, S610 to 670, and S710 to S760.

  The program can be provided through a network such as a website or e-mail. The program is stored in a recording medium such as a CD-ROM, DVD-ROM, CD-R, CD-RW, DVD-R, DVD-RW, MO, HDD, BD-ROM, BD-R, or BD-RE. It is also possible to record, save and distribute. This recording medium is read using a recording medium driving device, and the program code itself realizes the processing of the above embodiment, so that the recording medium also constitutes the present invention.

DESCRIPTION OF SYMBOLS 1 ... Moving means estimation model production | generation apparatus 10 ... GPS locus segment DB (1st database) without a label
20 ... GPS track segment DB with label (second database)
30 ... locus image extraction unit 40 ... unlabeled locus map image DB (third database)
50 ... Trace map image DB with label (4th database)
60 ... DNN learning unit 70 ... DNNDB
80 ... DNN feature quantity extraction unit 90 ... DNN feature quantity DB (fifth database)
100 ... basic feature quantity extraction unit 110 ... basic feature quantity DB (sixth database)
120... Feature amount connecting unit 130...
140 ... Estimated model generation unit 150 ... Estimated model DB
160 ... Map image DB (map image database)
170 ... Map image acquisition unit 180 ... Trajectory map image generation unit

Claims (7)

An apparatus for generating a moving means estimation model for estimating moving means based on a GPS trajectory,
A first database that stores, as GPS trajectory segments, enumeration information of positioning points acquired by GPS positioning for each section that is the same moving means;
A second database for storing labeled GPS trajectory segments with labels attached to the GPS trajectory segments;
A trajectory image extraction unit that visualizes the center of gravity of the positioning information of the positioning points when extracting a trajectory image from the positioning information of the positioning points of each GPS trajectory segment stored in the both databases;
A map image acquisition unit for acquiring a map image corresponding to each GPS trajectory segment stored in both databases from a map image database storing map images;
A trajectory extracted from the labeled GPS trajectory segment of the second database while generating an unlabeled trajectory map image based on the trajectory image extracted from the GPS trajectory segment of the first database and the map image corresponding to the GPS trajectory segment. A trajectory map image generation unit that generates a labeled trajectory map image based on the image and the map image corresponding to the labeled GPS trajectory segment;
A third database for storing the unlabeled locus map image;
A fourth database for storing the labeled locus map image in association with the label;
DNN feature values are extracted from learning results of the multilayer neural network based on the storage data of the third database and the fourth database, and the extracted DNN feature values are stored in the fifth database in correspondence with the labels. An extractor;
A basic feature amount extraction unit that extracts a basic feature amount from the labeled GPS trajectory segment of the second database and stores the extracted basic feature amount in the sixth database in association with the label;
A feature value linking unit that links DNN feature values and basic feature values corresponding to the fifth database and the sixth database, respectively, to be connected feature values;
An estimated model generation unit that generates the estimated model based on a seventh database that stores the connected feature values in association with the labels;
A moving means estimation model generation apparatus comprising: The map image acquisition unit
Means for calculating the longitude of the center-of-gravity coordinates of the GPS trajectory based on the GPS trajectory acquired from the both databases at regular time intervals;
Means for calculating the latitude of the barycentric coordinates of the GPS trajectory based on the GPS trajectory acquired at an interval of the predetermined time;
Means for obtaining from the map image database a map image included in a range of a longitude width and latitude width centered on the coordinates of the calculated longitude and latitude;
The moving means estimation model generation apparatus according to claim 1, comprising: The locus map image generation unit
The unlabeled trajectory map image is configured by combining a vector indicating a pixel element of the trajectory image extracted from the GPS trajectory segment of the first database and a vector indicating a pixel element of the map image corresponding to the GPS trajectory segment. Means for generating a vector indicating the elements of the pixel;
The labeled trajectory map image is obtained by combining a vector indicating a pixel element of the trajectory image extracted from the labeled GPS trajectory segment of the second database and a vector indicating a pixel of the map image corresponding to the labeled GPS trajectory segment. Means for generating a vector indicating the elements of the constituent pixels;
The moving means estimation model generation device according to claim 1, comprising: A first database that stores, as GPS trajectory segments, enumeration information of positioning points acquired by GPS positioning for each section that is the same moving means;
A second database for storing labeled GPS trajectory segments with labels attached to the GPS trajectory segments;
The computer generates a moving means estimation model for estimating the moving means, comprising:
A trajectory image extraction step for imaging the center of gravity of the positioning information of the positioning points when extracting the trajectory image from the positioning information of the positioning points of the GPS trajectory segments stored in the both databases;
A map image acquisition step for acquiring a map image corresponding to each GPS trajectory segment stored in the both databases from the map image database storing the map image;
A trajectory extracted from the labeled GPS trajectory segment of the second database while generating an unlabeled trajectory map image based on the trajectory image extracted from the GPS trajectory segment of the first database and the map image corresponding to the GPS trajectory segment. A trajectory map image generating step for generating a labeled trajectory map image based on the image and the map image corresponding to the labeled GPS trajectory segment;
A trajectory map image storage step of storing the unlabeled trajectory map image in a third database;
A labeled trajectory map image storing step of storing the labeled trajectory map image in a fourth database in association with the label;
DNN feature values are extracted from learning results of the multilayer neural network based on the storage data of the third database and the fourth database, and the DNN feature values are stored in the fifth database in correspondence with the labels. An extraction step;
A basic feature amount extracting step of extracting a basic feature amount from the labeled GPS trajectory segment of the second database and storing the extracted basic feature amount in the sixth database in association with the label;
A feature amount linking step of linking the corresponding DNN feature amounts and basic feature amounts of the fifth database and the sixth database, respectively, into a connected feature amount;
An estimated model generating step for generating the estimated model based on a seventh database storing the connected feature values in association with the labels;
A moving means estimation model generation method characterized by comprising: The map image acquisition step includes:
Calculating the longitude of the center-of-gravity coordinates of the GPS trajectory based on the GPS trajectory acquired from the both databases at regular time intervals;
Calculating the latitude of the barycentric coordinates of the GPS trajectory based on the GPS trajectory acquired at the predetermined time interval;
Obtaining a map image included in a range of a longitude width and a latitude width from the map image database centered on the coordinates of the calculated longitude and latitude; and
The moving means estimation model generation method according to claim 4, comprising: The locus map image generation step includes:
The unlabeled trajectory map image is configured by combining a vector indicating a pixel element of the trajectory image extracted from the GPS trajectory segment of the first database and a vector indicating a pixel element of the map image corresponding to the GPS trajectory segment. Generating a vector indicating the elements of the pixel;
The labeled trajectory map image is obtained by combining a vector indicating a pixel element of the trajectory image extracted from the labeled GPS trajectory segment of the second database and a vector indicating a pixel of the map image corresponding to the labeled GPS trajectory segment. Generating a vector indicating the elements of the constituent pixels;
The moving means estimation model generation method according to claim 4 or 5, characterized by comprising:   A moving means estimation model generation program that causes a computer to function as the moving means estimation model generation apparatus according to any one of claims 1 to 3.

Images