Rolling stock estimation device, platform door system, rolling stock estimation method and rolling stock estimation program

Abstract

To provide a technique capable of estimating a type of rolling stock entering a platform without changing the rolling stock.SOLUTION: A rolling stock estimation device 20A includes an image acquisition section 22A, an estimation section 24A and an output section 32. The image acquisition section 22A acquires a captured image including part of rolling stock entering a platform. The estimation section 24A estimates a type of the rolling stock included in the captured image by using an estimation model obtained by machine learning using the image including part of the rolling stock and rolling stock type information as teacher data. The estimation model is a model obtained by machine learning using images to which image processing for excluding predetermined feature information included in the images is applied as teacher data, as images including part of the rolling stock. The image acquisition section 22A acquires the image obtained by applying image processing to the captured image including part of the rolling stock, as a captured image. The output section 32 outputs information on the estimated rolling stock type.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a vehicle estimation device, a platform door system, a vehicle estimation method, and a vehicle estimation program. [Background technology]

[0002] In recent years, the installation of platform door devices has been promoted to further improve safety on platforms. Platform door devices have doors in different positions corresponding to the boarding and alighting doors of trains depending on the type of trains that stop at the platform. Therefore, it is necessary to distinguish the type of trains that enter the platform.

[0003] Patent Document 1 describes a method of arranging retroreflectors on a railway vehicle in a predetermined arrangement pattern according to the type of railway vehicle, and identifying the type of railway vehicle by photographing the retroreflectors. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2022-2955 Summary of the Invention [Problem to be solved by the invention]

[0005] The technology described in Patent Document 1 requires vehicles to be equipped with retroreflectors. To equip all vehicles with retroreflectors would require an increased number of man-hours. In addition, approval may be required to modify vehicles, which may result in procedural issues.

[0006] In view of the above problems, an object of the present invention is to provide a technology that can estimate the type of vehicle entering a platform without making any modifications to the vehicle. [Means for solving the problem]

[0007] In order to solve the above problems, a vehicle estimation device of one embodiment of the present invention includes an image acquisition unit that acquires a captured image including a part of a vehicle entering a platform, an estimation unit that estimates the type of vehicle included in the captured image using an estimation model trained by machine learning using the image including the part of the vehicle and vehicle type information as training data, and an output unit that outputs information on the estimated vehicle type, wherein the estimation model is a model trained by machine learning using, as training data, an image including the part of the vehicle, which has been subjected to image processing to remove certain feature information included in the image, and the image acquisition unit acquires, as the captured image, an image obtained by applying the image processing to the captured image including the part of the vehicle.

[0008] One embodiment of the present invention provides a platform door system comprising a vehicle estimation device and a platform door device, wherein the vehicle estimation device transmits information on the estimated vehicle type and information on the determined number of train formations to the platform door device, and the platform door device controls the opening and closing of the platform doors at a position corresponding to the train based on the information on the vehicle type and the information on the number of train formations.

[0009] A vehicle estimation method of one embodiment of the present invention includes the steps of acquiring a captured image including a part of a vehicle entering a platform, estimating the type of vehicle included in the captured image using an estimation model trained by machine learning using the image including the part of the vehicle and vehicle type information as training data, and outputting information on the estimated vehicle type, wherein the estimation model is a model trained by machine learning using, as training data, an image including the part of the vehicle, which has been subjected to image processing to remove certain feature information included in the image, and the step of acquiring the captured image is a step of acquiring, as the captured image, an image obtained by applying the image processing to an image captured including a part of the vehicle.

[0010] A vehicle estimation program of one embodiment of the present invention is a program for causing a computer to execute the steps of acquiring a photographed image including a portion of a vehicle entering a platform, estimating the type of vehicle of the vehicle included in the photographed image using an estimation model trained by machine learning using the image including the portion of the vehicle and vehicle type information as training data, and outputting information on the estimated vehicle type, wherein the estimation model is a model trained by machine learning using, as training data, an image including part of the vehicle, which has been subjected to image processing to remove certain feature information included in the image, and the step of acquiring the photographed image is a step of acquiring, as the photographed image, an image obtained by applying the image processing to an image including part of the vehicle.

[0011] In addition, any combination of the above, or mutual substitution of the components or expressions of the present invention among methods, devices, programs, temporary or non-temporary storage media recording programs, systems, etc., are also valid aspects of the present invention. Effect of the Invention

[0012] According to the present invention, the type of vehicle entering a platform can be estimated without making any modifications to the vehicle. [Brief description of the drawings]

[0013] [Figure 1] FIG. 2 is a functional block diagram of the platform door system according to the first embodiment. [Diagram 2] FIG. 2 is a diagram illustrating an example of an installation position of the imaging device illustrated in FIG. [Diagram 3] FIG. 3 is a diagram showing a schematic front or rear view of the vehicle shown in FIG. 2. [Figure 4] FIG. 1 is a block diagram showing a schematic diagram of a first configuration example of a platform door system. [Diagram 5] FIG. 2 is a block diagram illustrating a second configuration example of a platform door system. [Figure 6] FIG. 11 is a functional block diagram of a platform door system according to a second embodiment. [Figure 7] FIG. 3 is a diagram showing a schematic side view of the vehicle shown in FIG. 2. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Among the embodiments disclosed in this specification, those that are composed of multiple objects may be integrated, and conversely, those that are composed of one object may be divided into multiple objects. Regardless of whether they are integrated or not, it is sufficient that they are configured to achieve the object of the invention.

[0015] Among the embodiments disclosed in this specification, those in which multiple functions are provided in a distributed manner may have some or all of the multiple functions integrated together, and conversely, those in which multiple functions are provided in a distributed manner may have some or all of the multiple functions integrated together. Regardless of whether the functions are integrated or distributed, it is sufficient that the configuration is such that the object of the invention can be achieved.

[0016] In addition, separate components that have something in common are distinguished by adding "first," "second," etc. to the beginning of their names, and these are omitted when referring to them collectively. In addition, terms including ordinal numbers such as first and second are used to describe various components, but these terms are used only for the purpose of distinguishing one component from other components, and do not limit the components.

[0017] Hereinafter, the present disclosure will be described based on preferred embodiments with reference to the drawings. In the embodiments and modified examples, the same or equivalent components and members are given the same reference numerals, and duplicated descriptions are omitted as appropriate. In addition, the dimensions of the members in each drawing are enlarged or reduced as appropriate for ease of understanding. In addition, some of the members that are not important for explaining the embodiments are omitted in each drawing.

[0018] [First embodiment] Hereinafter, the first embodiment will be described with reference to the drawings. FIG. 1 is a functional block diagram of a platform door system 1 according to the first embodiment. Each functional block shown in each drawing including FIG. 1 can be realized in hardware terms by electronic elements and mechanical parts such as a computer CPU, and in software terms by a computer program, but here, functional blocks realized by the cooperation of these are depicted. Therefore, it will be understood by those skilled in the art that these functional blocks can be realized in various ways by combining hardware and software.

[0019] As shown in Fig. 1, the platform door system 1 includes a plurality of imaging devices 10, a vehicle estimation device 20, and a platform door device 60. Although Fig. 1 shows three imaging devices 10 as the plurality of imaging devices 10, the plurality of imaging devices 10 may be two imaging devices 10, or four or more imaging devices 10. Hereinafter, when describing matters common to each of the plurality of imaging devices 10, they will be simply referred to as imaging devices 10 as appropriate.

[0020] The multiple imaging devices 10 are installed at different positions on the platform. The imaging device 10 includes, for example, a known optical camera. The imaging device 10 captures an image including at least a part of a vehicle, such as a railroad car, entering the platform. The imaging device 10 may always capture an image toward a position where the vehicle is captured when the vehicle enters the platform, or may capture an image in response to a signal that detects the entry of the vehicle. In the following, an example in which the imaging direction and zoom magnification are constant so that the imaging device 10 captures an image with a constant angle of view will be described, but the imaging direction and zoom magnification of the imaging device 10 may be variable. In the following, it will be described that the imaging device 10 captures an image from the time it receives a signal that detects the entry of the vehicle until just before the vehicle completely stops, but the timing of capturing the image is not limited.

[0021] Fig. 2 is a diagram showing a schematic diagram of an example of an installation position of the imaging device 10. Fig. 2 shows imaging devices 10a to 10c as an example of a plurality of imaging devices 10. The imaging device 10a is installed near the end of the platform 80 on the side in the traveling direction 86 of the train 92. The imaging device 10a can capture an image including at least a part of the front of the leading car 90a of the train 92 and at least a part of the side on the platform 80 side.

[0022] In the example shown in FIG. 2, train 92 has cars 90a, 90b, and 90c coupled in this order from the front in the traveling direction 86. Train 92 is not limited to a three-car formation, and may be a one-car formation, a two-car formation, or a four-car or more formation. Furthermore, when train 92 includes multiple cars 90, the multiple cars 90 may be of the same model, or may be a mixture of different models. Hereinafter, the side of train 92 in traveling direction 86 will also be referred to as the "front," and the side opposite to traveling direction 86 will also be referred to as the "rear."

[0023] The imaging device 10b is installed near the center of the platform 80 in the traveling direction 86 of the train 92. The imaging device 10b is capable of capturing an image including at least a portion of the side surface of the second car 90b from the front of the train 92 facing the platform 80. The imaging device 10c is installed near the end of the platform 80 on the opposite side to the traveling direction 86 of the train 92. The imaging device 10c is capable of capturing an image including at least a portion of the rear surface of the last car 90c of the train 92 and at least a portion of the side surface facing the platform 80.

[0024] The imaging device 10 may be installed in a position where passengers are unlikely to be captured in the image. For example, the imaging device 10 may be installed so as to be suspended from the ceiling of the platform 80. For example, the imaging device 10 may be installed in an area 82 between the platform door device 60 and a train 92. Here, the external configuration of the platform door device 60 shown in Fig. 2 includes a plurality of doors 66 that can be opened and closed, and a plurality of door pockets 68 in which each of the plurality of doors 66 can be stored.

[0025] The multiple imaging devices 10 may include an imaging device 10 that photographs the train 92 from the front and an imaging device 10 that photographs the train 92 from the rear. The imaging device 10a is an example of an imaging device 10 that photographs the train 92 from the front. The imaging device 10c is an example of an imaging device 10 that photographs the train 92 from the rear.

[0026] The multiple imaging devices 10 may capture images including at least a part of the vehicle 90 at positions corresponding to the multiple stop target positions on the platform 80. Here, the stop target position is a position indicating at which position on the platform 80 the leading vehicle 90 of the train 92 should stop, and differs depending on the type of vehicle 90 constituting the train 92 and the number of train formations of the train 92. In addition, the stop target positions are set in advance at multiple positions on the platform 80. Therefore, the multiple imaging devices 10 capture images including at least a part of the vehicle 90 at positions corresponding to the multiple stop target positions on the platform 80, respectively, so that even if the stop target positions of the train 92 differ depending on the number of train formations of the train 92, etc., the image including the leading vehicle 90 of the train 92 can be appropriately captured. Note that, instead of the multiple imaging devices 10, one imaging device 10 capable of simultaneously capturing images of the vehicle 90 at positions corresponding to the multiple stop target positions on the platform 80 may be used. Here, the number of trains in the train 92 is information indicating the number of cars 90 included in the train 92 in the train.

[0027] FIG. 3 is a diagram showing a schematic view of the front or rear of a vehicle 90. As shown in FIG. 2 and FIG. 3, the imaging device 10 may capture at least one of a departure sign (not shown) installed on the platform 80 and a destination display 94 displayed on the vehicle 90. The departure sign is a guide display device that displays information such as the departure time of the train 92, the destination of the train 92, and the type of vehicle 90 that constitutes the train 92, and is, for example, an electronic bulletin board. The destination display 94 may be displayed on the front of the leading vehicle 90a of the train 92, and may be captured by the imaging device 10a. Also, for example, the destination display 94 may be displayed on the rear of the last vehicle 90c of the train 92, and may be captured by the imaging device 10c.

[0028] As shown in FIG. 1, the vehicle estimation device 20 includes an image acquisition unit 22, an estimation unit 24, a character recognition unit 26, a discrimination unit 28, a comprehensive estimation unit 30, and an output unit 32.

[0029] The image acquiring unit 22 acquires captured images captured by the multiple imaging devices 10. In other words, the image acquiring unit 22 acquires captured images including a portion of the vehicle 90 entering the platform 80 from multiple positions. The captured image including a portion of the vehicle 90 may include at least a portion of the vehicle 90, and may include the entire vehicle 90.

[0030] The image acquiring unit 22 may acquire, for example, the following captured images according to the above-mentioned installation locations of the multiple imaging devices 10. For example, the image acquiring unit 22 may acquire a captured image of a train 92 including vehicles 90 captured from the front and a captured image of the train 92 captured from the rear. For example, the image acquiring unit 22 may acquire captured images corresponding to multiple target stopping positions on the platform 80. The image acquiring unit 22 may acquire a captured image including at least one of a departure board installed on the platform 80 and a destination display displayed on the vehicle 90.

[0031] The estimation unit 24 uses the estimation model to estimate the vehicle type of the vehicle 90 for each of the images captured from multiple positions acquired by the image acquisition unit 22. The estimation model is a model that has been machine-trained using images including a part of the vehicle 90 and vehicle type information as training data.

[0032] Here, the estimation model used by the estimation unit 24 will be described. Since the train 92 runs on the track 84 (see FIG. 2), it is likely to appear at the same position in the angle of view of the image captured by the imaging device 10. For this reason, it is preferable to use an image classification model as the estimation model rather than an object detection model. The machine learning method may be any supervised learning method, and for example, a Convolutional Neural Network (CNN) may be used.

[0033] Among the teacher data of the estimation model, images including a part of the vehicle 90 can be collected by any method, but may be images captured using a plurality of imaging devices 10 of the platform door system 1, for example. In this case, in order to reduce the influence of changes in the surrounding environment, images of the vehicle 90 captured at various times over a certain period of time, for example, one year, may be used. This makes it possible to reduce the influence of changes in scenery due to seasonal changes, changes in the position of the sun, changes in weather, and the like. Among the teacher data of the estimation model, the vehicle type information may be input by the creator of the estimation model as the vehicle type of the vehicle 90 included in each of the collected images.

[0034] When the image acquisition unit 22 acquires an image of a train 92 including a vehicle 90 taken from the front and an image taken from the rear, the estimation unit 24 may estimate the vehicle types of the front and rear vehicles included in the image taken from the front and the image taken from the rear. Here, the front and rear vehicles mean that the front vehicle is coupled forward of the rear vehicle. In the example shown in FIG. 2, the image taken from the front by the imaging device 10a includes the vehicle 90a as the front vehicle, and the image taken from the rear by the imaging device 10c includes the vehicle 90c as the rear vehicle.

[0035] The overall estimation unit 30 estimates information on the vehicle type based on the estimation result by the estimation unit 24. In detail, the estimation unit 24 estimates the vehicle type of the vehicle 90 for each of a plurality of captured images taken from a plurality of positions when a certain train 92 enters the platform 80. The overall estimation unit 30 estimates information on one vehicle type by integrating information on the vehicle type of the vehicle 90 estimated by the estimation unit 24 for each of a plurality of captured images. For example, the overall estimation unit 30 may estimate the vehicle type estimated by the estimation unit 24 for the most captured images as the vehicle type of the vehicle 90. In addition, for example, the overall estimation unit 30 may change the weighting of the estimation result by the estimation unit 24 for each captured image. Specifically, the estimation unit 24 outputs a score according to the certainty of the result together with information on the vehicle type estimated for each captured image. The overall estimation unit 30 may estimate the vehicle type with the highest score as the vehicle type of the vehicle 90 by tallying up the scores for each estimated vehicle type.

[0036] The character recognition unit 26 recognizes character information included in the captured image acquired by the image acquisition unit 22. Specifically, when the captured image acquired by the image acquisition unit 22 includes at least one of a departure board installed on the platform and a destination display 94 displayed on the vehicle 90, as described above, the character recognition unit 26 recognizes character information included in the departure board or the destination display 94. For character recognition by the character recognition unit 26, a known character recognition technology such as OCR (Optical Character Recognition) may be used.

[0037] The discrimination unit 28 discriminates the number of train formations of the train 92 including the vehicles 90 based on at least one photographed image acquired by the image acquisition unit 22. For example, a specific imaging device 10 is installed so as to capture the entire train 92 in one photographed image. The discrimination unit 28 may discriminate the number of train formations for one photographed image by the imaging device 10 based on the area occupied by the train 92 in the photographed image. Also, for example, a plurality of imaging devices 10 are installed so as to capture images of different positions along the traveling direction 86 of the train 92. The discrimination unit 28 may discriminate the number of train formations for the images captured by each of the plurality of imaging devices 10 by specifying the installation positions of one or more imaging devices 10 that captured an image including at least a part of the train 92.

[0038] The output unit 32 outputs information on the type of vehicle 90 that is comprehensively estimated by the comprehensive estimation unit 30 based on the estimation result by the estimation unit 24. The output unit 32 may further output character information recognized by the character recognition unit 26 together with the information on the type of vehicle 90. The output unit 32 may further output information on the number of formations determined by the determination unit 28 together with the information on the type of vehicle 90. The output unit 32 transmits this information to the platform door device 60. The transmission of information from the output unit 32 to the platform door device 60 may use wired communication or wireless communication. Publicly known communication technologies may be used for these communications.

[0039] The platform door device 60 includes a control panel 62 and a plurality of door control units 64. The control panel 62 controls each of the plurality of door control units 64. Each of the plurality of door control units 64 corresponds to one door 66, and controls the opening and closing of the corresponding door 66 in response to a control signal from the control panel 62.

[0040] The control panel 62 identifies a door 66 at a position corresponding to the train 92 based on the information on the type of vehicle 90 and the information on the number of formations of the train 92 received from the vehicle estimation device 20. After that, the control panel 62 transmits an opening / closing instruction to the door control unit 64 corresponding to the door 66 at the identified position in response to receiving an opening / closing operation instruction signal operated by the conductor or the like.

[0041] The control panel 62 may store in advance information on the position of the corresponding door 66 according to a combination of information on the type of vehicle 90 and information on the number of formations of the train 92. The control panel 62 may store in advance a condition for the position of the door 66 according to the information on the type of vehicle 90 and a condition for the position of the door 66 according to the information on the number of formations of the train 92, and specify the position of the door 66 that satisfies both of these conditions. The same applies when it is estimated that there are multiple types of vehicles 90 included in the train 92.

[0042] In this way, the platform door device 60 controls the opening and closing of doors at a position corresponding to the train 92 based on information on the type of vehicle 90 and information on the number of trains 92 in formation.

[0043] As described above, the platform door system 1 includes a plurality of imaging devices 10, a vehicle estimation device 20, and a platform door device 60. Each of these components included in the platform door system 1 may be configured as a single device, or may be configured as a plurality of devices. FIG. 4 is a block diagram that shows a schematic diagram of a first configuration example of the platform door system 1. As shown in FIG. 4, the vehicle estimation device 20 may be configured to include a plurality of edge devices 34 and a computer 36. In this case, each of the plurality of edge devices 34 includes an image acquisition unit 22 and an estimation unit 24. The computer 36 includes an overall estimation unit 30 and an output unit 32. The character recognition unit 26 and the discrimination unit 28 may be included in the edge device 34 or the computer 36. According to such a configuration example, the edge device 34 corresponding to each imaging device 10 performs processing such as estimation processing on the captured image, and therefore the processing can be distributed.

[0044] FIG. 5 is a block diagram showing a schematic diagram of a second configuration example of the platform door system 1. As shown in FIG. 6, the vehicle estimation device 20 may be configured to include a GPU (Graphics Processing Unit) server 38 and a computer 36. In this case, the GPU server 38 includes the image acquisition unit 22 and the estimation unit 24. The computer 36 includes the overall estimation unit 30 and the output unit 32. The character recognition unit 26 and the discrimination unit 28 may be included in the GPU server 38 or in the computer 36. According to this configuration example, the GPU server 38 collectively performs processes such as estimation processes on the images captured by the imaging devices 10, so that the processes can be centralized.

[0045] The configurations and effects of the present embodiment described above are as follows.

[0046] The vehicle estimation device 20 according to this embodiment includes an image acquisition unit 22, an estimation unit 24, and an output unit 32. The image acquisition unit 22 acquires captured images including a part of a vehicle 90 entering a platform 80 from a plurality of positions. The estimation unit 24 estimates the type of vehicle 90 for each of the captured images from the plurality of positions using an estimation model trained by machine learning using the images including the part of the vehicle 90 and vehicle type information as teacher data. The output unit 32 outputs information on the overall estimated type of vehicle 90 based on the estimation result by the estimation unit 24.

[0047] This allows vehicle estimation device 20 to estimate the type of vehicle 90 entering platform 80, without the need to make any changes to vehicle 90 or to acquire information such as operation information of vehicle 90. Furthermore, vehicle estimation device 20 estimates the type of vehicle 90 for each of images captured from a plurality of positions, and outputs information on the type of vehicle 90 that is comprehensively estimated based on the estimation results, thereby improving the accuracy of the estimation.

[0048] In this embodiment, the image acquisition unit 22 acquires a photographed image of a train 92 including vehicles 90 taken from the front and a photographed image of the train 92 taken from the rear. The estimation unit 24 estimates the types of the front and rear vehicles 90 included in each of the two acquired photographed images. This allows the vehicle estimation device 20 to estimate the types of the respective vehicles 90 even if the train 92 entering the platform 80 is a train consisting of vehicles 90 of different types coupled to each other at the front and rear.

[0049] In this embodiment, the image acquisition unit 22 acquires captured images corresponding to a plurality of stop target positions on the platform 80. This allows the vehicle estimation device 20 to appropriately acquire captured images of the leading vehicle 90 of the train 92 even if the stop target positions of the train 92 differ depending on the number of train formations of the train 92, etc.

[0050] In this embodiment, the image acquisition unit 22 further acquires a captured image including at least one of a departure sign installed on the platform 80 and a destination display 94 displayed on the vehicle 90. The vehicle estimation device 20 further includes a character recognition unit 26 that recognizes character information included in the captured image. The output unit 32 further outputs the character information. The character information may directly or indirectly indicate information on the vehicle type of the vehicle 90. Therefore, the vehicle estimation device 20 can output the character information as information that can be used to auxiliary determine whether or not the information on the estimated vehicle type of the vehicle 90 was correct.

[0051] The vehicle estimation device 20 according to this embodiment further includes a discrimination unit 28 that discriminates the number of formations of a train 92 including vehicles 90 based on at least one captured image. The output unit 32 further outputs information on the discriminated number of formations. As a result, the vehicle estimation device 20 outputs information on the number of formations of the train 92 in addition to information on the type of vehicle 90, and therefore can output more detailed information about the train 92.

[0052] The platform door system 1 according to this embodiment includes a vehicle estimation device 20 and a platform door device 60. The vehicle estimation device 20 transmits information on the estimated vehicle type of the vehicle 90 and information on the determined number of train formations of the train 92 to the platform door device 60. The platform door device 60 controls the opening and closing of doors 66 at a position corresponding to the train 92, based on the information on the vehicle type of the vehicle 90 and the information on the number of train formations of the train 92. This allows the platform door system 1 to adjust the opening positions of the doors 66 of the platform door device 60 in accordance with the positions of the boarding and alighting gates, which differ depending on the vehicle type of the vehicle 90 and the number of train formations of the train 92.

[0053] [Second embodiment] Next, a second embodiment will be described with reference to the drawings. In the drawings and description of the second embodiment, components and members that are the same as or equivalent to those in the first embodiment are given the same reference numerals. Explanations that overlap with the first embodiment will be omitted as appropriate, and the description will focus on the configurations that differ from the first embodiment.

[0054] Fig. 6 is a functional block diagram of a platform door system 1A according to the second embodiment. As shown in Fig. 6, the platform door system 1A includes a plurality of imaging devices 10, a vehicle estimation device 20A, and a platform door device 60. The plurality of imaging devices 10 and the platform door device 60 are similar to those in the first embodiment, and therefore description thereof will be omitted.

[0055] The vehicle estimation device 20A includes an image processing unit 40, an image acquisition unit 22A, an estimation unit 24A, a character recognition unit 26, a discrimination unit 28, a comprehensive estimation unit 30, and an output unit 32. The character recognition unit 26, the discrimination unit 28, the comprehensive estimation unit 30, and the output unit 32 are similar to those in the first embodiment, and therefore descriptions thereof will be omitted. In this embodiment, the vehicle estimation device 20A does not need to include the comprehensive estimation unit 30, and in that case, the output unit 32 may output the estimation result by the estimation unit 24A.

[0056] The image processing unit 40 acquires an image captured by the imaging device 10 and performs image processing to remove predetermined feature information contained in the image. The image acquisition unit 22A acquires, as a captured image, the image that has been subjected to image processing by the image processing unit 40. Other functions of the image acquisition unit 22A are similar to those of the image acquisition unit 22 in the first embodiment.

[0057] The estimation unit 24A estimates the type of vehicle 90 included in the captured image acquired by the image acquisition unit 22A using the estimation model. The estimation model is a model that has been machine-learned using an image including a part of the vehicle 90 and vehicle type information as training data. Of the training data of the estimation model of this embodiment, an image including a part of the vehicle 90 is an image that has been subjected to image processing to remove predetermined feature information included in the image. The image processing to remove the predetermined feature information here is similar to the image processing that the image processing unit 40 executes on the image captured by the imaging device 10. In this embodiment, the image that the image processing unit 40 applies the above-mentioned image processing to the image captured by the imaging device 10 is described as being collected as an image of training data for the estimation model. The machine learning method used in the estimation model of this embodiment, the period for collecting training data, the method for collecting vehicle type information, and the like are similar to the estimation model of the first embodiment, so description thereof will be omitted.

[0058] Here, the background of image processing by the image processing unit 40 will be described. In terms of appearance, the color and pattern of the vehicle body surface of the vehicle 90 are likely to appear as major features. Therefore, when an image of the vehicle 90 is directly used as teacher data, the estimation result of the vehicle type of the vehicle 90 is likely to depend on whether the color and pattern of the vehicle body surface of the vehicle 90 are similar or not. However, even if the vehicle 90 is the same model, the paint pattern may be renewed or a special specification such as a wrapped vehicle may be applied. In such a case, the color and pattern of the vehicle body surface of the vehicle 90 changes significantly, so that it may be difficult to estimate the vehicle type of the vehicle 90. In addition, the color and pattern of the vehicle body surface of the vehicle 90 are likely to change depending on the weather and time of day at the time of shooting. For example, in weather such as snow, rain, and fog, or in a time such as night, information on the color and pattern of the vehicle body surface of the vehicle 90 is likely to be lost from the captured image.

[0059] Therefore, in this embodiment, an image that has been subjected to image processing to remove predetermined feature information is used as an image of teacher data for the estimation model used by the estimation unit 24A. Also, the image acquisition unit 22A acquires, as a captured image, an image that has been subjected to image processing on an image captured including a part of the vehicle 90. This reduces the influence of feature information that is likely to change depending on the conditions when the vehicle 90 is captured, and improves the accuracy of estimation.

[0060] The image processing performed by the image processing unit 40 may include a process of reducing color information contained in the image. The process of reducing color information may be, for example, grayscaling. This reduces the proportion of the information on the color and pattern of the vehicle body surface of the vehicle 90 that occupies a characteristic portion of the captured image, making it easier to estimate the vehicle 90 based on elements other than the color and pattern of the vehicle body surface.

[0061] The image processing performed by the image processing unit 40 may include masking of specific regions of the image. In this case, for example, the image processing unit 40 may extract from the image a feature region to be used for estimating the vehicle type other than the color and pattern of the vehicle body surface of the vehicle 90, and perform masking of the region other than the feature region as a specific region. The feature region may be, for example, the outline of the entire vehicle 90, the driver's window 98 (see FIG. 3), or the lights 96 (see FIG. 3) on the front and back of the vehicle 90. FIG. 7 is a diagram showing a schematic side of the vehicle 90. The feature region may be an entrance / exit 100 on the side of the vehicle 90 and a window 102 on the side of the vehicle 90 shown in FIG. 7. These regions are usually regions that are unlikely to change even in the case of a vehicle 90 with special specifications such as a wrapped vehicle. In addition, information such as the shape, arrangement, and number of these feature regions may be used as a feature for estimating the vehicle type of the vehicle 90.

[0062] The image processing performed by the image processing unit 40 may include both the above-mentioned color information reduction process and mask process. In addition, the image processing performed by the image processing unit 40 may be other image processing as long as it is image processing that removes predetermined feature information contained in the image.

[0063] The configurations and effects of the present embodiment described above are as follows: Description of the same configurations and effects as those of the first embodiment will be omitted.

[0064] The vehicle estimation device 20A according to the present embodiment includes an image acquisition unit 22A, an estimation unit 24A, and an output unit 32. The image acquisition unit 22A acquires a captured image including a part of a vehicle 90 entering a platform 80. The estimation unit 24A estimates the type of vehicle 90 included in the captured image using an estimation model trained by machine learning using an image including a part of the vehicle 90 and vehicle type information as training data. The estimation model is a model trained by machine learning using an image including a part of the vehicle 90, which has been subjected to image processing to remove predetermined feature information included in the image, as training data. The image acquisition unit 22 acquires an image obtained by performing image processing on an image including a part of the vehicle 90 as a captured image. The output unit 32 outputs information on the estimated vehicle type.

[0065] This allows the vehicle estimation device 20A to estimate the type of vehicle 90 entering the platform 80, without the need to make any changes to the vehicle 90 or to acquire information such as operation information of the vehicle 90. Furthermore, the vehicle estimation device 20A estimates the type of vehicle 90 included in a captured image that has been subjected to similar image processing, using an estimation model that has been machine-learned using as training data an image that has been subjected to image processing to remove predetermined feature information. Therefore, the estimation accuracy can be improved by reducing the influence of feature information that is easily changed depending on the conditions when the vehicle 90 is photographed, from the estimation result of the type of vehicle 90.

[0066] In this embodiment, the image processing includes a process of reducing color information contained in the image. This allows the vehicle estimation device 20A to reduce color information, which is a characteristic part that is particularly likely to change depending on the shooting conditions in an image including the vehicle 90, and therefore the accuracy of estimating the type of vehicle 90 can be further improved.

[0067] In this embodiment, the image processing includes a mask process for a specific region of the image. This allows the vehicle estimation device 20A to remove a region that is particularly likely to change depending on the shooting conditions in the image including the vehicle 90 from the captured image, thereby further improving the accuracy of estimating the type of vehicle 90.

[0068] Above, examples of the embodiments of the present invention have been described in detail. The above-mentioned embodiments merely show specific examples of implementing the present invention. The contents of the embodiments do not limit the technical scope of the present invention, and many design changes such as changing, adding, and deleting components are possible within the scope of the invention as defined in the claims. In the above-mentioned embodiments, the contents for which such design changes are possible are described with the notation "of the embodiment" or "in the embodiment", but design changes may also be permitted for contents without such notation.

[0069] Any combination of the above-described embodiments and modifications is also useful as an embodiment of the present invention. A new embodiment resulting from the combination has the combined effects of the respective embodiments and modifications.

[0070] Among the embodiments disclosed in this specification, those in which multiple functions are provided in a distributed manner may have some or all of the multiple functions integrated together, and conversely, those in which multiple functions are provided in a distributed manner may have some or all of the multiple functions integrated together. Regardless of whether the functions are integrated or distributed, it is sufficient that the configuration is such that the object of the invention can be achieved. [Explanation of symbols]

[0071] 1...platform door system, 10...imaging device, 20...vehicle estimation device, 22...image acquisition unit, 24...estimation unit, 26...character recognition unit, 28...discrimination unit, 32...output unit, 40...image processing unit, 60...platform door device, 66...door, 80...platform, 90...vehicle, 92...train, 94...destination display.

Claims

1. an image acquisition unit that acquires a photographed image including a part of a vehicle entering a platform; an estimation unit that estimates the type of vehicle included in the captured image by using an estimation model that has been machine-learned using an image including a part of the vehicle and vehicle type information as training data; an output unit that outputs information about the estimated vehicle type; Equipped with The estimation model is a model trained by machine learning using as training data an image that includes a part of the vehicle and that has been subjected to image processing to remove predetermined feature information included in the image, The image acquisition unit acquires, as the captured image, an image obtained by subjecting a captured image including a part of the vehicle to the image processing. Vehicle estimator.

2. The image processing includes a process of reducing color information contained in the image. The vehicle estimation device according to claim 1 .

3. The image processing includes a mask process for a specific region of the image. The vehicle estimation device according to claim 1 .

4. The image acquisition unit acquires the captured images from a plurality of positions, The estimation unit estimates a type of a vehicle included in each of the images captured from the plurality of positions; The output unit outputs information on the vehicle type comprehensively estimated based on the estimation result. The vehicle estimation device according to claim 1 .

5. A discrimination unit that discriminates the number of train formations including the vehicle based on the captured image, The output unit also outputs information on the determined number of trains. The vehicle estimation device according to any one of claims 1 to 4.

6. A platform door system comprising the vehicle estimation device according to claim 5 and a platform door device, the vehicle estimation device transmits information on the estimated vehicle type and information on the determined number of train formations to the platform door device; The platform door device controls opening and closing of platform doors at a position corresponding to the train based on information on the vehicle type and information on the number of train formations. Platform door system.

7. acquiring a photographed image including a part of a vehicle entering a platform; estimating a vehicle type of the vehicle included in the captured image using an estimation model that has been machine-learned using an image including a part of the vehicle and vehicle type information as training data; outputting information on the estimated vehicle type; Including, The estimation model is a model trained by machine learning using as training data an image that includes a part of the vehicle and that has been subjected to image processing to remove predetermined feature information included in the image, The step of acquiring the captured image is a step of acquiring, as the captured image, an image obtained by subjecting a captured image including a part of the vehicle to the image processing. Vehicle estimation method.

8. acquiring a photographed image including a part of a vehicle entering a platform; estimating a vehicle type of the vehicle included in the captured image using an estimation model that has been machine-learned using an image including a part of the vehicle and vehicle type information as training data; outputting information on the estimated vehicle type; A vehicle estimation program for causing a computer to execute the above, The estimation model is a model trained by machine learning using as training data an image that includes a part of the vehicle and that has been subjected to image processing to remove predetermined feature information included in the image, The step of acquiring the captured image is a step of acquiring, as the captured image, an image obtained by subjecting a captured image including a part of the vehicle to the image processing. Vehicle estimation program.

Images