Traffic flow analysis device, traffic flow analysis method and program

Abstract

[Problem] To improve analysis accuracy when analyzing people and vehicles that constitute a traffic flow. [Solution] This traffic flow analysis device comprises: an acquiring means that acquires images from a camera installed in a location enabling image capture of mobile bodies subject to traffic flow analysis; a storing means that stores a plurality of types of identification methods for identifying attributes of the mobile bodies captured by the camera; a selecting means that selects an identification method suited to the trend of the mobile bodies captured by the camera from among the plurality of types of identification methods stored in the storing means; and an identifying means that identifies the captured mobile bodies in the acquired images and the attributes thereof using the identification method selected by the selecting means.

Description

[Technical Field]

[0001] The present invention relates to a traffic flow analysis device, a traffic flow analysis method, and a program recording medium. [Background technology]

[0002] Patent Document 1 discloses a people flow prediction device that can robustly predict people flow despite spatial changes. According to this document, the people flow prediction device selects a prediction model from the model storage means based on prediction conditions including a prediction period to be predicted and tolerance conditions related to the characteristics of a prediction model created in advance. The people flow prediction device then predicts people flow data under the prediction conditions based on the selected prediction model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2021 / 130926 Summary of the Invention [Problem to be solved by the invention]

[0004] Patent Document 1 describes the prediction of people flow using a prediction model, but does not mention detailed analysis of traffic flow including people and vehicles, and in particular, does not mention measures to improve the accuracy of such analysis.

[0005] An object of the present invention is to provide a traffic flow analysis device, a traffic flow analysis method, and a program recording medium that can improve the accuracy of analyzing people and vehicles that make up a traffic flow. [Means for solving the problem]

[0006] According to a first aspect, there is provided a traffic flow analysis device comprising: an acquisition means for acquiring images from a camera installed in a position where it can photograph a moving object that is the subject of traffic flow analysis; a storage means for storing a plurality of identification methods for identifying the attributes of the moving object captured by the camera; a selection means for selecting an identification method that matches the tendency of the moving object captured by the camera from the plurality of identification methods stored in the storage means; and an identification means for identifying the moving object and its attributes captured in the acquired image using the identification method selected by the selection means.

[0007] According to a second aspect, there is provided a traffic flow analysis method, which includes selecting an identification method that matches the tendency of the moving object captured by the camera from among a plurality of identification methods stored in a storage means that stores the identification methods for identifying the attributes of the moving object captured by the camera, acquiring an image from the camera, and identifying the attributes of the moving object captured in the acquired image using the selected identification method.

[0008] According to a third aspect, there is provided a program recording medium that causes a computer to execute the following processes: selecting an identification method that matches the tendency of the moving object captured by the camera from among a plurality of identification methods stored in a storage means that stores the identification methods for identifying the attributes of the moving object captured by the camera, which is installed in a position where the moving object that is the subject of traffic flow analysis can be photographed; acquiring an image from the camera; and identifying the attributes of the moving object captured in the acquired image using the selected identification method. [Effects of the Invention]

[0009] According to the present invention, a traffic flow analysis device, a traffic flow analysis method, and a program recording medium are provided that can improve the accuracy of analyzing people and vehicles that make up traffic flows. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a diagram showing a configuration of an embodiment of the present invention; [Figure 2]3 is a flow diagram illustrating the operation of one embodiment of the present invention. [Figure 3] FIG. 2 is a diagram for explaining the operation of one embodiment of the present invention. [Figure 4] 1 is a diagram illustrating a configuration of a traffic flow analysis device according to a first embodiment of the present invention. [Figure 5] 1 is a diagram showing an example of the arrangement of a traffic flow analysis device and cameras according to a first embodiment of the present invention. [Figure 6] FIG. 3 is a diagram illustrating a classification model selection operation according to the first exemplary embodiment of the present invention. [Figure 7] 3 is a flowchart showing the operation of the traffic flow analysis device according to the first embodiment of the present invention. [Figure 8] FIG. 4 is a diagram illustrating a configuration of a traffic flow analysis device according to a second embodiment of the present invention. [Figure 9] FIG. 10 is a diagram illustrating a classification model selection operation according to the second exemplary embodiment of the present invention. [Figure 10] 5 is a flowchart showing the operation of the traffic flow analysis device according to the second embodiment of the present invention. [Figure 11] FIG. 10 is a diagram illustrating a configuration of a traffic flow analysis device according to a third embodiment of the present invention. [Figure 12] FIG. 10 is a diagram illustrating a classification model selection operation according to the third exemplary embodiment of the present invention. [Figure 13] 10 is a flowchart showing the operation of the traffic flow analysis device according to the third embodiment of the present invention. [Figure 14] FIG. 10 is a diagram illustrating a configuration of a traffic flow analysis device according to a fourth embodiment of the present invention. [Figure 15] FIG. 10 is a diagram showing an example of the layout of facilities around an intersection where cameras are installed in a fourth embodiment of the present invention. [Figure 16] FIG. 10 is a diagram illustrating a classification model selection operation according to the fourth exemplary embodiment of the present invention. [Figure 17] 10 is a flowchart showing the operation of the traffic flow analysis device according to the fourth embodiment of the present invention. [Figure 18] FIG. 10 is a diagram illustrating a configuration of a traffic flow analysis device according to a fifth embodiment of the present invention. [Figure 19]FIG. 13 is a diagram illustrating a classification model selection operation according to the fifth exemplary embodiment of the present invention. [Figure 20] FIG. 1 is a diagram showing the configuration of a computer that can function as a traffic flow analysis device of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] First, an overview of one embodiment of the present invention will be described with reference to the drawings. Note that the reference numerals in this overview are used for convenience to identify each element as an example to facilitate understanding, and are not intended to limit the present invention to the illustrated form. Furthermore, connecting lines between blocks in the drawings and the like referred to in the following description include both bidirectional and unidirectional lines. Unidirectional arrows are used to schematically indicate the flow of main signals (data) and do not exclude bidirectionality. A program is executed via a computer device, which includes, for example, a processor, a storage device, an input device, a communication interface, and, if necessary, a display device. Furthermore, this computer device is configured to be able to communicate with internal or external devices (including computers) via the communication interface, whether wired or wireless. Furthermore, ports or interfaces are present at the input / output connection points of each block in the drawings, but are not shown.

[0012] In one embodiment, the present invention can be realized by a traffic flow analysis device 10 connected to one or more cameras 500a, 500b, as shown in Figure 1. More specifically, the traffic flow analysis device 10 includes an acquisition means 11, a storage means 14, a selection means 13, and an identification means 12.

[0013] The acquisition means 11 acquires images from cameras 500a and 500b installed in positions where they can capture images of moving objects that are the subject of traffic flow analysis.

[0014] The storage means 14 stores a plurality of types of identification methods for identifying the attributes of moving objects captured by the cameras 500a and 500b. Simply put, the storage means 14 stores classification models and processing algorithms for identifying the gender and age group of people captured by the cameras 500a and 500b. The classification models can be created by various types of machine learning using training data that associates images of people captured by the cameras 500a and 500b with attribute information (ground truth data) about those people. The processing algorithms can be optimized based on the trends in the flow of people expected to be captured by the cameras 500a and 500b.

[0015] The selection means 13 selects an identification method that matches the tendency of the moving object captured by the cameras 500a, 500b from the plurality of identification methods stored in the storage means 14. The identification means 12 uses the identification method selected by the selection means 13 to identify the moving object captured in the acquired image and its attribute (class).

[0016] In the above configuration, the identification methods stored in the traffic flow analysis device 10 are created in multiple types based on the tendencies of moving objects captured by the cameras 500a and 500b, which have been investigated in advance, and the selection means 13 selects an identification method from the multiple types of identification methods using selection rules that are determined to suit the tendencies of moving objects captured by the cameras 500a and 500b.

[0017] Next, a traffic flow analysis method used in the traffic flow analysis device 10 of this embodiment will be described in detail with reference to the drawings. Fig. 2 is a flowchart showing the operation of the traffic flow analysis device 10. First, the traffic flow analysis device 10 selects an identification method from the plurality of identification methods (step S001). At this time, the traffic flow analysis device 10 selects an identification method using a selection rule that is determined to select an identification method that matches the tendency of moving objects captured by the cameras 500a and 500b.

[0018] Next, the traffic flow analysis device 10 acquires image data from the cameras 500a and 500b (step S002). The traffic flow analysis device 10 uses the selected identification method to identify moving objects and their attribute information from the acquired image data (step S003). Note that in the example of FIG. 2, the moving objects and their attribute information are identified together, but the identification of the moving objects and the identification of their attributes may be performed separately. In this case, the traffic flow analysis device 10 detects moving objects from frame differences and the like that make up the image, extracts the image of the detected moving object, and performs the identification process.

[0019] Here, the operation of selecting an identification method using the above selection rules will be described. FIG. 3 is a diagram for explaining the operation of one embodiment of the present invention. For example, assume that camera 500a is installed in an area where, according to past statistical data, there is a high volume of female traffic. On the other hand, assume that camera 500b is installed in an area where, according to past statistical data, there is a high volume of male traffic. In this case, for analyzing image data acquired from camera 500a, the selection means 13 selects an identification method suitable for an area where there is a high volume of female traffic. This can improve the accuracy of identifying moving objects and their attribute information from the image data acquired from camera 500a. Similarly, for analyzing image data acquired from camera 500b, the selection means 13 selects an identification method suitable for an area where there is a high volume of male traffic. This can improve the accuracy of identifying moving objects and their attribute information from the image data acquired from camera 500b.

[0020] As described above, according to this embodiment, the traffic flow analysis device 10 selects an identification method that matches the tendency of moving objects captured by the cameras 500a and 500b to analyze the traffic flow, thereby improving the accuracy of analyzing people and vehicles that make up the traffic flow.

[0021] Furthermore, in the above explanation, the identification method is selected, but as one aspect of the identification method, the same effect can be obtained by changing the classification model or processing algorithm for identifying moving objects.

[0022] In the above description, the identification method is changed depending on the camera 500a, 500b that captured the image data. However, the identification method can also be changed depending on other conditions. For example, if the analysis target changes from people to vehicles, the identification method can be changed. Furthermore, if a camera is installed at an intersection, the identification method can be changed depending on the traffic light or other factors related to the analysis target, such as analyzing people when the light is green and analyzing vehicles when the light is red. The identification method can also be changed depending on the analysis time period related to the analysis target, such as analyzing people from morning to night and analyzing vehicles from late at night to dawn. Of course, the conditions for selecting these identification methods can be defined as selection rules, and the traffic flow analysis device 10 can select an identification method by referring to the selection rules.

[0023] In addition, in the above-described embodiment, an example has been described in which the moving body is a person, but the moving body is not limited to a person. For example, the moving body may be a car, a bicycle or other light vehicle, a UAV (Unmanned Aerial Vehicle), an automatic guided vehicle, etc.

[0024] In the above embodiment, it has been described that a plurality of types of identification methods for identifying attributes of moving objects are pre-stored in the storage means 14. However, it is desirable that these identification methods be added or updated to the most appropriate ones as appropriate. For example, classification models and processing algorithms with higher prediction accuracy are set in the storage means 14 periodically or when an event occurs, and the selection means 13 selects the most appropriate ones. Possible examples of the periodic addition or update include once a month or once a week. Possible examples of the event include the release or version upgrade of the classification model processing algorithm, as well as changes in the required accuracy of attribute identification due to legal amendments or changes in safety levels.

[0025] [First embodiment] Next, a first embodiment of the present invention in which a classification model is selected as an example of the identification method will be described in detail with reference to the drawings. In the following description, an example in which a classification model is selected using a model selection rule that selects a classification model based on the position of a camera will be described.

[0026] 4 is a diagram showing the configuration of a traffic flow analysis device according to the first embodiment of the present invention. The traffic flow analysis device 100 includes an acquisition unit 101, a model storage unit 104, a model selection unit 103, and an identification unit 102.

[0027] The acquisition means 101 acquires images from cameras 500a and 500b installed in positions where they can capture images of moving objects that are the subject of traffic flow analysis.

[0028] The model storage means 104 stores a plurality of classification models for identifying the attributes of moving objects captured by the cameras 500a and 500b. Simply put, the model storage means 104 stores classification models for identifying the gender and age group of people captured by the cameras 500a and 500b. Such classification models can be created by various types of machine learning using training data that associates images of people captured by the cameras 500a and 500b with attribute information (ground truth data) about those people.

[0029] The model selection means 103 selects a classification model from the model storage means 104. The identification means 102 uses the classification model selected by the model selection means 103 to identify the moving object and its attribute (class) shown in the acquired image.

[0030] In the above configuration, the classification models stored in the traffic flow analysis device 100 are created in multiple types based on the tendencies of moving objects captured by the cameras 500a and 500b, which have been investigated in advance, and the model selection means 103 selects a classification model from the multiple types of classification models using model selection rules that are determined to match the tendencies of moving objects captured by the cameras 500a and 500b.

[0031] Fig. 5 is a diagram showing an example of the arrangement of the traffic flow analysis device 100 and cameras 500a and 500b in the first embodiment of the present invention. As shown in Fig. 5, in the first embodiment, the camera 500a is set on the north side of the station STA and is capable of capturing images of passersby. On the other hand, the camera 500b is set on the south side of the station STA and is also capable of capturing images of passersby. The traffic flow analysis device 100 is capable of acquiring images captured by the cameras 500a and 500b and analyzing the flow of people.

[0032] The configuration of the traffic flow analysis device 100 is almost the same as that shown in FIG. 1, and therefore a detailed description thereof will be omitted. FIG. 6 is a diagram illustrating the classification model selection operation of this embodiment. The model storage means 104 of this embodiment stores multiple classification models for identifying the attributes of pedestrians captured by the cameras 500a and 500b. Such traffic flow analysis devices 100 may be installed for each area or intersection. For example, a traffic flow analysis device 100 installed in each area or near that area may acquire and analyze images from cameras installed in the area it is responsible for. Similarly, a traffic flow analysis device 100 installed at each intersection or near that intersection may acquire and analyze images from cameras installed at that intersection. Furthermore, such a traffic flow analysis device 100 may be one or more MEC servers installed at the edge of a network and transmit analysis results to an information processing device installed on the cloud side. Note that "MEC" stands for Multi-access Edge Computing or Mobile Edge Computing.

[0033] As shown in FIG. 6, in this embodiment, the model storage means 104 stores classification model A and classification model B. Classification model A is a classification model tuned for identifying the attributes of pedestrians photographed by camera 500a. Such classification models can be created by performing machine learning using images actually photographed by camera 500a (or images photographed at a similar position) and actual people flow data (ground truth data) as training data. Similarly, classification model B is a classification model tuned for identifying the attributes of pedestrians photographed by camera 500b.

[0034] Then, the model selection means 103 of this embodiment selects a classification model using a model selection rule that selects an appropriate classification model depending on the installation positions of the cameras 500 a and 500 b. Specifically, the model selection means 103 selects classification model A for images captured by the camera 500 a, and selects classification model B for images captured by the camera 500 b.

[0035] Next, the operation of this embodiment will be described in detail with reference to the drawings. Fig. 7 is a flowchart showing the operation of the traffic flow analysis device 100 of the first embodiment of the present invention. Referring to Fig. 7, first, the traffic flow analysis device 100 selects a classification model for each camera and sets it in the identification means 102 (step S101).

[0036] Thereafter, the traffic flow analysis device 100 repeats the process of detecting pedestrians captured in the images taken by the cameras 500a and 500b and identifying their attributes (step S102).

[0037] As described above, the model selection means 103 selects an appropriate classification model according to the installation positions of the cameras 500a and 500b. For example, as shown in FIG. 6, since there are many adults at the installation position of the camera 500a, classification model A specialized for adults is selected. Since pedestrians with various attributes pass by at the installation position of the camera 500b, classification model B specialized for identifying these attributes is selected. This enables the traffic flow analysis device 100 to accurately identify the attributes of pedestrians.

[0038] [Second embodiment] Next, a second embodiment of the present invention will be described in detail with reference to the drawings, in which a classification model is selected using a model selection rule that selects a classification model based on the time period when an image was captured. FIG. 8 is a diagram showing the configuration of a traffic flow analysis device 100a according to the second embodiment of the present invention. The difference from the traffic flow analysis device of the first embodiment shown in FIG. 2 is that the model storage means 114 stores classification models for each time period, and the model selection means 113 selects a classification model according to the time period during which analysis is performed in addition to the installation positions of the cameras 500a and 500b. Since the other configurations are the same as those of the first embodiment, their explanations will be omitted and the following description will focus on the differences.

[0039] As shown in FIG. 9 , in this embodiment, the model storage means 114 stores a plurality of classification models for different time periods, such as classification model A0810 to classification model A1416. Classification model A0810 is a classification model tuned for identifying the attributes of pedestrians photographed by camera 500a between 8:00 and 10:00. Similarly, classification model A1416 is a classification model tuned for identifying the attributes of pedestrians photographed by camera 500a between 14:00 and 16:00. Such classification models can be created by performing machine learning using images actually photographed by camera 500a in each time period (or images photographed at similar positions) and actual people flow data (ground truth data) as training data.

[0040] 9 shows classification models for two time periods, 8:00-10:00 and 14:00-16:00, but classification models may be held for other time periods as well, or one classification model may be used for multiple time periods. Also, in the example of Fig. 9, the classification model for camera 500b is omitted, but classification models for each time period may also be prepared for camera 500b.

[0041] Next, the operation of this embodiment will be described in detail with reference to the drawings. Fig. 10 is a flowchart showing the operation of the traffic flow analysis device 100a according to the second embodiment of the present invention. Referring to Fig. 10, first, the traffic flow analysis device 100a selects a classification model corresponding to the time period for which analysis is to be performed for each camera, and sets the selected classification model in the identification means 102 (step S201).

[0042] Thereafter, the traffic flow analysis device 100a repeats the process of detecting pedestrians captured in the images taken by the cameras 500a and 500b and identifying their attributes (step S202). As mentioned above, the model selection means 113 selects an appropriate classification model according to the installation location and time period of the camera 500a. For example, as shown in Fig. 9, even for the same camera 500a, since there are many commuters and adults between 8:00 and 10:00 in the morning, classification model A0810 specialized for that time period is selected. On the other hand, since children and elderly people also pass by between 14:00 and 16:00 in the afternoon, classification model A1416 specialized for that time period is selected.

[0043] According to this embodiment, which operates as described above, it is possible to improve the accuracy of identifying pedestrian attributes in places where trends may change depending on the time of day. This is because a classification model that takes time of day into consideration is prepared and selected.

[0044] [Third embodiment] Next, a third embodiment of the present invention will be described in detail with reference to the drawings. This embodiment dynamically selects a classification model using model selection rules that select the classification model based on the trends of attributes identified by the identification means. FIG. 11 is a diagram showing the configuration of a traffic flow analysis device 100b according to the third embodiment of the present invention. The difference from the traffic flow analysis device of the first embodiment shown in FIG. 2 is that the model storage means 124 stores classification models for different pedestrian flow trends, and the model selection means 123 selects a classification model based on the most recent pedestrian flow analysis results in addition to the installation positions of the cameras 500a and 500b. Since the other configurations are the same as those of the first embodiment, their explanations will be omitted and the following description will focus on the differences.

[0045] As shown in FIG. 12, in this embodiment, the model storage means 124 stores a plurality of classification models for different pedestrian flow tendencies, such as classification model a to classification model x. Classification model a is a classification model tuned for identifying pedestrian attributes in a situation where there are only adult males. Similarly, classification model b is a classification model tuned for identifying pedestrian attributes in a situation where there is a large variation in attributes. Classification model c is a classification model tuned for identifying pedestrian attributes in a situation where there are an equal number of adult males and adult females passing by.

[0046] Next, the operation of this embodiment will be described in detail with reference to the drawings. Fig. 13 is a flowchart showing the operation of the traffic flow analysis device 100b according to the third embodiment of the present invention. Referring to Fig. 13, first, the traffic flow analysis device 100b acquires the most recent analysis results for each of the cameras 500a and 500b (step S301). The most recent analysis results may be those held by the traffic flow analysis device 100b, or may be those managed by a higher-level device that receives the analysis results from the traffic flow analysis device 100b.

[0047] Next, the traffic flow analysis device 100b selects a classification model that is closest to the most recent analysis results for each of the cameras 500a and 500b, and sets it in the identification means 102 (step S302). For example, if the most recent analysis results for the image from camera 500a show a high proportion of adult males, the model selection means 123 selects classification model a, which is tuned for identifying pedestrian attributes in a situation where there are a high proportion of adult males. Similarly, if the most recent analysis results for the image from camera 500b show a high degree of attribute variation, the model selection means 123 selects classification model b.

[0048] Thereafter, the traffic flow analysis device 100b repeats the process of detecting pedestrians captured in the images taken by the cameras 500a and 500b and identifying their attributes (step S303).

[0049] According to this embodiment, which operates as described above, it is possible to further improve the accuracy of identifying pedestrian attributes compared to Embodiment 1. The reason for this is that it employs a configuration in which multiple types of classification models for different targets are prepared and a classification model is selected based on the trends in people flow obtained from the most recent analysis.

[0050] [Fourth embodiment] Next, a fourth embodiment of the present invention will be described in detail with reference to the drawings. This embodiment selects a classification model using a model selection rule that selects a classification model based on the colors of traffic signal lights installed around the camera. FIG. 14 is a diagram showing the configuration of a traffic flow analysis device 100c according to the fourth embodiment of the present invention. The first difference from the traffic flow analysis device of the first embodiment shown in FIG. 2 is that a traffic light information acquisition means 135 is added. The second difference is that a model storage means 134 stores classification models for each color of traffic signal lights, and the model selection means 133 selects a classification model based on the colors of the traffic signal lights in addition to the installation positions of the cameras 500a and 500b. Since the other configurations are the same as those of the first embodiment, their explanations will be omitted and the following description will focus on the differences.

[0051] The signal light information acquisition means 135 acquires the color of the lights of the traffic signals at the intersections where the cameras 500a and 500b are installed. The signal light information acquisition means 135 can acquire the color of the lights of the traffic signals by acquiring signal control information from a signal control device that controls the traffic signals, or by determining the color from the color of the lights of the traffic signals captured in the images captured by the cameras 500a and 500b.

[0052] Fig. 15 is a diagram showing an example of the layout of facilities around an intersection where camera 500a is installed. In the example of Fig. 15, there is an office ahead of pedestrian traffic light SIG1, and an elementary school ahead of pedestrian traffic light SIG2. Therefore, when pedestrian traffic light SIG1 turns green, pedestrians coming and going from the office are photographed by camera 500a. After that, when pedestrian traffic light SIG2 turns green, children attending the elementary school are photographed by camera 500a.

[0053] As shown in FIG. 16 , the model storage means 134 of this embodiment stores two classification models: classification model ASIG1 and classification model ASIG2. Classification model ASIG1 is a classification model tuned for identifying attributes of pedestrians photographed while the pedestrian traffic light SIG1 is green. Similarly, classification model ASIG2 is a classification model tuned for identifying attributes of pedestrians photographed while the pedestrian traffic light SIG2 is green. Such classification models can be created by performing machine learning using images actually photographed by the camera 500a at the timing of each traffic light (or images photographed at similar positions) and actual pedestrian flow data (ground truth data) as training data. Note that the example of FIG. 16 shows classification models applied when the pedestrian traffic lights SIG1 and SIG2 are both green, but classification models for other light states may also be stored. Furthermore, in the examples of FIGS. 15 and 16, the classification model for camera 500b is omitted, but a classification model for each color of traffic signal light may also be prepared for camera 500b in the same manner.

[0054] Next, the operation of this embodiment will be described in detail with reference to the drawings. Fig. 17 is a flowchart showing the operation of the traffic flow analysis device 100c according to the fourth embodiment of the present invention. Referring to Fig. 17, first, the traffic flow analysis device 100c acquires the color of the light of a pedestrian traffic light for each camera (step S401).

[0055] Next, the traffic flow analysis device 100c selects a classification model corresponding to the color of the light of the pedestrian traffic signal for each camera, and sets it in the identification means 102 (step S402).

[0056] Thereafter, the traffic flow analysis device 100c repeats the process of detecting pedestrians in the images captured by the cameras 500a and 500b using the selected classification model and identifying their attributes (step S403).

[0057] As described above, the model selection means 133 selects an appropriate classification model according to the installation position of the camera 500a and the color of the pedestrian traffic light. For example, as shown in Fig. 16, when the pedestrian traffic light SIG1 is green at the intersection where the camera 500a is installed, the classification model ASIG1 specialized for that period is selected. On the other hand, when the pedestrian traffic light SIG2 is green at the intersection where the camera 500a is installed, the classification model ASIG2 specialized for that period is selected.

[0058] According to this embodiment, which operates as described above, it is possible to further improve the accuracy of identifying pedestrian attributes compared to embodiment 1. The reason for this is that a classification model is prepared and selected that takes into account not only the camera position but also the color of traffic lights, which affect pedestrian flow.

[0059] [Fifth embodiment] Next, a fifth embodiment of the present invention, which analyzes the attributes of vehicles as traffic flows, will be described in detail with reference to the drawings. FIG. 18 is a diagram showing the configuration of a traffic flow analysis device 100d according to the fifth embodiment of the present invention. The difference from the traffic flow analysis device of the first embodiment shown in FIG. 2 is that a classification model for vehicles is stored in the model storage means 144, and the model selection means 143 selects a classification model according to the installation positions of the cameras 500a and 500b. Since the other configurations are the same as those of the first embodiment, their explanations will be omitted and the following description will focus on the differences.

[0060] As shown in FIG. 19 , in this embodiment, the model storage means 144 stores classification models VA and VB. The classification model VA is a classification model tuned for identifying the attributes of vehicles captured by the camera 500a. Such a classification model can be created by performing machine learning using an image actually captured by the camera 500a (or an image captured at a similar position) and actual traffic flow data (ground truth data). Similarly, the classification model VB is a classification model tuned for identifying the attributes of vehicles captured by the camera 500b. Note that the vehicles may include bicycles, electric kick scooters, and other light vehicles.

[0061] Then, the model selection means 143 of this embodiment selects a classification model using a model selection rule that selects an appropriate classification model depending on the installation positions of the cameras 500 a and 500 b. Specifically, the model selection means 143 selects classification model VA for the image captured by camera 500 a, and selects classification model VB for the image captured by camera 500 b.

[0062] 7, the traffic flow analysis device 100d selects a classification model for each camera and sets it in the identification means 102 (step S101). Thereafter, the traffic flow analysis device 100d detects pedestrians captured in images captured by the cameras 500a and 500b and repeats the process of identifying their attributes (step S102).

[0063] As described above, the present invention can also be applied to a traffic flow analysis device that detects vehicles and identifies their attributes. In addition, like the second to fourth embodiments of the first embodiment, this embodiment can be modified or expanded to a form in which classification models are prepared according to time periods, analysis results of recent traffic flows, and the colors of traffic signal lights, and these can be selected.

[0064] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and further modifications, substitutions, and adjustments can be made without departing from the basic technical concept of the present invention. For example, the network configurations, element configurations, and data representation formats shown in the drawings are examples intended to aid in understanding the present invention, and the present invention is not limited to the configurations shown in these drawings.

[0065] For example, in the above-described embodiments, the analysis targets are pedestrians and vehicles, but the analysis targets are not limited to these. The analysis targets may also be limited by a specific gender, age, whether or not a person has a disability, etc.

[0066] Furthermore, the above-described embodiments are merely examples, and the traffic flow analysis device 100 can be configured to select an identification method such as a classification model or a processing algorithm based on various conditions. For example, the traffic flow analysis device 100 can select a classification model based on any combination of conditions such as the location where the image was taken, the time period, recent trends, etc.

[0067] (About hardware configuration) In each embodiment of the present disclosure, each component of each device represents a functional block. Some or all of the components of each device are realized by any combination of an information processing device 900 and a program, for example, as shown in Fig. 20. Fig. 20 is a block diagram showing an example of the hardware configuration of the information processing device 900 that realizes each component of each device. The information processing device 900 includes, as an example, the following configuration. ·CPU(Central Processing Unit)901 ROM (Read Only Memory) 902 ·RAM(Random Access Memory)903 Program 904 loaded into RAM 903 A storage device 905 for storing a program 904 A drive device 907 for reading and writing the program recording medium 906 A communication interface 908 for connecting to a communication network 909 Input / output interface 910 for inputting and outputting data Bus 911 connecting each component

[0068] Each component of each device in each embodiment is realized by the CPU 901 acquiring and executing a program 904 that realizes the function. That is, the CPU 901 in FIG. 20 executes a vehicle detection program and a determination program, and performs an update process for each calculation parameter stored in the RAM 903, the storage device 905, or the like. The program 904 that realizes the function of each component of each device is stored in advance in a program recording medium such as the storage device 905 or the ROM 902, and is read out by the CPU 901 as needed. The program 904 may be supplied to the CPU 901 via the communication network 909, or may be stored in advance in the program recording medium 906, and the drive device 907 may read out the program and supply it to the CPU 901.

[0069] Furthermore, this program 904 can display the processing results, including intermediate states, at each stage as necessary on a display device, or can communicate with the outside via a communication interface. Furthermore, this program 904 can be recorded on a computer-readable (non-transitive) program recording medium.

[0070] There are various variations in the method of realizing each device. For example, each device may be realized by any combination of a separate information processing device 900 and a program for each component. Furthermore, multiple components included in each device may be realized by any combination of a single information processing device 900 and a program. That is, the communication terminals and network control devices shown in the first to third embodiments, and processors installed in these devices, can be realized by computer programs that cause the above-mentioned processes to be executed using the hardware.

[0071] In addition, some or all of the components of each device may be realized by other general-purpose or dedicated circuits, processors, etc., or a combination of these. These may be configured by a single chip, or by multiple chips connected via a bus.

[0072] Some or all of the components of each device may be realized by a combination of the above-mentioned circuits and programs.

[0073] When some or all of the components of each device are realized by multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or decentralized. For example, the information processing devices, circuits, etc. may be realized as a client-server system, a cloud computing system, or the like, in a form in which each device is connected via a communication network.

[0074] It should be noted that the above-described embodiments are preferred embodiments of the present disclosure, and the scope of the present disclosure is not limited to only the above-described embodiments. In other words, those skilled in the art can modify or substitute the above-described embodiments to construct various modified forms without departing from the gist of the present disclosure.

[0075] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.

[0076] [Appendix 1] an acquisition means for acquiring images from a camera installed in a position capable of photographing a moving object that is a target of traffic flow analysis; a storage means for storing a plurality of types of identification methods for identifying attributes of moving objects captured by the camera; a selection means for selecting an identification method that matches the tendency of a moving object captured by the camera from the plurality of identification methods stored in the storage means; an identification means for identifying a moving object and its attributes shown in the acquired image by using the identification method selected by the selection means; Traffic flow analysis device. [Appendix 2] The selection means of the traffic flow analysis device may be configured to select the identification method based on the position of the camera. [Appendix 3] The selection means of the traffic flow analysis device may be configured to select the identification method based on the time period when the image was taken. [Appendix 4] The selection means of the traffic flow analysis device may be configured to select the identification method based on the tendency of the attribute identified by the identification means. [Appendix 5] The selection means of the traffic flow analysis device may be configured to select the identification method based on the color of the light of a traffic signal installed around the camera. [Appendix 6] The identification method selected by the traffic flow analysis device may be a classification model or a processing algorithm for identifying attributes of moving objects captured by a camera. [Appendix 7] In the above traffic flow analysis device, a plurality of types of the classification model or processing algorithm are created based on trends of moving objects captured by the camera that have been investigated in advance; The selection means can be configured to select the classification model or processing algorithm from the plurality of classification models or processing algorithms using selection rules that are defined to select a classification model or processing algorithm that matches the tendency of moving objects captured by the camera. [Appendix 8] In the above traffic flow analysis device, The moving object may be a person, and the classification model may be a classification model created by machine learning using statistical data on people flow at the location where the camera is installed as training data. [Appendix 9] In the above traffic flow analysis device, The moving object may be a person, and the classification model may be created by machine learning using statistical data of people flow by time period captured by the camera as training data. [Appendix 10] selecting an identification method that matches the tendency of the moving object captured by the camera from among a plurality of identification methods stored in a storage means for storing the plurality of identification methods for identifying the attributes of the moving object captured by the camera installed at a position where the moving object to be analyzed by traffic flow can be captured; acquiring an image from the camera; A traffic flow analysis method that uses the selected identification method to identify attributes of moving objects captured in the acquired images. [Appendix 11] a process of selecting an identification method that matches the tendency of a moving object captured by a camera from among a plurality of identification methods stored in a storage means for storing the plurality of identification methods for identifying attributes of a moving object captured by a camera installed at a position where the moving object to be analyzed can be captured; acquiring an image from the camera; and a program recording medium storing a program for causing a computer to execute a process of identifying attributes of a moving object shown in the acquired image using the selected identification method. The embodiments of Supplementary Notes 10 and 11 can be expanded to the embodiments of Supplementary Notes 2 to 9, similarly to Supplementary Note 1.

[0077] The disclosures of the above-cited patent documents are incorporated herein by reference and may be used as the basis or part of the present invention, as necessary. Modifications and adjustments of the embodiments and examples are possible within the scope of the entire disclosure of the present invention (including the claims), and further based on its basic technical concept. Furthermore, various combinations and selections (including partial deletions) of the various disclosed elements (including each element of each claim, each element of each embodiment or example, each element of each drawing, etc.) are possible within the scope of the disclosure of the present invention. In other words, the present invention naturally embraces various modifications and alterations that would be possible by a person skilled in the art in accordance with the entire disclosure and technical concept, including the claims. In particular, with regard to the numerical ranges set forth herein, any numerical value or subrange within that range should be construed as specifically set forth, even if not otherwise specified. Furthermore, the disclosures of the above-cited documents, when used in part or in whole in combination with the disclosures herein as part of the disclosure of the present invention, in accordance with the spirit of the present invention, are also deemed to be included in the disclosures of this application. [Explanation of symbols]

[0078] 500a, 500b camera 10, 100, 100a, 100b, 100c, 100d traffic flow analyzer 11, 101 Acquisition means 12, 102 Identification means 13 Selection methods 14 Memory means 103, 113, 123, 133, 143 Model selection measures 104, 114, 124, 134, 144 Model storage means 135 Signal light information acquisition means 900 Information Processing Equipment 901 CPU(Central Processing Unit) 902 ROM (Read Only Memory) 903 RAM (Random Access Memory) 904 Program 905 Storage device 906 Program recording media 907 Drive unit 908 Communication Interface 909 Communication Network 910 Input / Output Interface 911 Bus SIG1, SIG2 pedestrian traffic lights STA Station

Claims

1. an acquisition means for acquiring images from a camera installed in a position capable of photographing a moving object that is a target of traffic flow analysis; a storage means for storing a plurality of types of identification methods for identifying attributes of moving objects captured by the camera; a selection means for selecting an identification method that matches the tendency of a moving object captured by the camera from the plurality of identification methods stored in the storage means; an identification means for identifying a moving object and its attributes in the acquired image by using the identification method selected by the selection means, The selection means selects the identification method based on at least one of the position of the camera, the tendency of the attribute identified by the identification means, and the color of the light of a traffic signal installed around the camera. Traffic flow analysis device.

2. An acquisition means for acquiring images from a camera installed in a position capable of photographing a moving object that is the subject of traffic flow analysis; a storage means for storing a plurality of types of identification methods for identifying attributes of moving objects captured by the camera; a selection means for selecting an identification method that matches the tendency of a moving object captured by the camera from the plurality of identification methods stored in the storage means; an identification means for identifying a moving object and its attributes in the acquired image by using the identification method selected by the selection means, the identification method is a classification model or processing algorithm for identifying attributes of moving objects captured by the camera; a plurality of types of the classification model or processing algorithm are created based on trends of moving objects captured by the camera that have been investigated in advance; the selection means selects the classification model or processing algorithm from the plurality of classification models or processing algorithms using a selection rule that is set to select a classification model or processing algorithm that matches the tendency of the moving object captured by the camera; The moving object is a person, and the classification model is a classification model created by machine learning using statistical data on people flow at the position where the camera is installed as training data. Traffic flow analysis device.

3. 3. The traffic flow analysis device according to claim 1, wherein the selection means selects the identification method based on a time period when the image was taken.

4. A computer comprising: selecting an identification method that matches the tendency of the moving object captured by the camera from among a plurality of identification methods stored in a storage means for storing the plurality of identification methods for identifying the attributes of the moving object captured by the camera installed at a position where the moving object to be analyzed by traffic flow can be captured; acquiring an image from the camera; Identifying attributes of the moving object captured in the acquired image using the selected identification method; The identification method is selected based on at least one of the location of the camera, the tendency of the identified attribute, and the color of the light of a traffic signal installed around the camera. Traffic flow analysis method.

5. A computer comprising: selecting an identification method that matches the tendency of the moving object captured by the camera from among a plurality of identification methods stored in a storage means for storing the plurality of identification methods for identifying the attributes of the moving object captured by the camera installed at a position where the moving object to be analyzed by traffic flow can be captured; acquiring an image from the camera; Identifying attributes of the moving object captured in the acquired image using the selected identification method; the identification method is a classification model or processing algorithm for identifying attributes of moving objects captured by the camera; a plurality of types of the classification model or processing algorithm are created based on trends of moving objects captured by the camera that have been investigated in advance; the selection of the identification method comprises selecting a classification model or a processing algorithm from the plurality of classification models or processing algorithms using a selection rule that is set to select a classification model or a processing algorithm that matches the tendency of the moving object captured by the camera; The moving object is a person, and the classification model is a classification model created by machine learning using statistical data on people flow at the position where the camera is installed as training data. Traffic flow analysis method.

6. a process of selecting an identification method that matches the tendency of a moving object captured by a camera from among a plurality of identification methods stored in a storage means for storing the plurality of identification methods for identifying attributes of a moving object captured by a camera installed at a position where the moving object to be analyzed can be captured; acquiring an image from the camera; and causing the computer to execute a process of identifying attributes of the moving object captured in the acquired image using the selected identification method; The process of selecting the identification method is a process of selecting the identification method based on at least one of the position of the camera, the tendency of the identified attribute, and the color of the light of a traffic signal installed around the camera. program.

7. A process of selecting an identification method that matches the tendency of a moving object captured by a camera from among a plurality of identification methods stored in a storage means for storing the identification methods for identifying the attributes of a moving object captured by a camera installed in a position where the moving object can be photographed as the subject of traffic flow analysis; acquiring an image from the camera; and causing the computer to execute a process of identifying attributes of the moving object captured in the acquired image using the selected identification method; the identification method is a classification model or processing algorithm for identifying attributes of moving objects captured by the camera; a plurality of types of the classification model or processing algorithm are created based on trends of moving objects captured by the camera that have been investigated in advance; the process of selecting an identification method is a process of selecting a classification model or a processing algorithm from the plurality of classification models or processing algorithms using a selection rule that is defined to select a classification model or a processing algorithm that is suited to a tendency of a moving object captured by the camera, The moving object is a person, and the classification model is a classification model created by machine learning using statistical data on people flow at the position where the camera is installed as training data. program.

Images