Obstacle detection device and obstacle detection system

Abstract

The present invention provides an obstacle detection device and an obstacle detection system capable of accurately detecting obstacles around platform doors based on captured images. [Solution] The obstacle detection device 100 according to the present disclosure is an obstacle detection device 100 connected to an imaging device 300 that captures an image of the area around the platform door 200 so as to include the platform door 200, and is characterized by comprising: an image acquisition unit 102 that acquires an image from the imaging device 300; an obstacle detection processing unit 103 that has a larger data processing capacity than the imaging device 300, detects the presence or absence of an obstacle 500 around the platform door 200 based on the image, and generates obstacle information that indicates at least the presence or absence of an obstacle 500 around the platform door 200; and an obstacle detection controller 104 that outputs obstacle information to an individual control panel 201 that controls the opening and closing of the platform door 200.

Description

Technical Field

[0001] The present disclosure relates to an obstacle detection device and an obstacle detection system.

Background Art

[0002] Conventionally, detection of obstacles around a home door has been performed by an obstacle sensor attached to the home door. The obstacle sensor consists of a planar sensor, which transmits scanning light composed of infrared rays or the like, and receives and measures the reflected light from an object to determine the presence or absence of an obstacle. However, such an obstacle sensor has a problem that false detection occurs due to factors such as weather and the reflectivity of the object. Therefore, in recent years, a technique has been proposed in which, instead of the above-described obstacle sensor, an obstacle is detected by processing an image captured by a camera or the like.

[0003] In the invention described in Patent Document 1, a detection area set around a home door is imaged by an AI camera, and an obstacle is detected based on the captured image.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the invention described in Patent Document 1, the data capacity that can be processed by the AI camera is not considered. Therefore, in the invention described in Patent Document 1, there is a problem that the data capacity that can be processed by the AI camera itself as an imaging device is small, and accurate obstacle detection cannot be sufficiently performed.

[0006] This disclosure is made to solve the problems described above, and aims to provide an obstacle detection device and an obstacle detection system that can accurately detect obstacles around platform doors based on captured images. [Means for solving the problem]

[0007] The obstacle detection device according to this disclosure is an obstacle detection device connected to an imaging device that captures images of the area around a platform door, including the platform door, and is characterized by comprising: an image acquisition unit that acquires images from the imaging device; an obstacle detection processing unit that has a larger data processing capacity than the imaging device, detects the presence or absence of obstacles around the platform door based on the images, and generates obstacle information that indicates at least the presence or absence of obstacles around the platform door; and an obstacle detection controller that outputs obstacle information to an individual control panel that controls the opening and closing of the platform door.

[0008] Furthermore, the obstruction detection system according to this disclosure is characterized by comprising the above-mentioned obstruction detection device, individual control panels installed on the platform doors, and an integrated control unit that controls the platform doors in a comprehensive manner by controlling the individual control panels. [Effects of the Invention]

[0009] According to this disclosure, the obstacle detection device and obstacle detection system can accurately detect obstacles around platform doors based on captured images. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing the configuration of the obstacle detection system according to Embodiment 1. [Figure 2] This is a schematic diagram showing a platform screen door according to Embodiment 1. [Figure 3] This is a schematic diagram illustrating the processing of the obstacle detection processing unit in Embodiment 1. [Figure 4] This is a diagram showing the configuration of the learning device related to the obstacle detection device of Embodiment 1. [Figure 5]This is a diagram showing the configuration of the obstacle detection device according to Embodiment 1. [Figure 6] This is a schematic diagram showing an example of training image data in Embodiment 1. [Figure 7] Block diagram showing hardware configuration examples for each configuration of Embodiment 1. [Figure 8] This is a flowchart showing the processing flow of the obstacle detection system according to Embodiment 1. [Figure 9] This is a block diagram showing the configuration of the obstacle detection system according to Embodiment 2. [Figure 10] This is a block diagram showing the configuration of the obstacle detection system according to Embodiment 3. [Figure 11] This is a block diagram showing the configuration of the obstacle detection system according to Embodiment 4. [Figure 12] This is a schematic diagram showing an example of training image data for Embodiment 4. [Figure 13] This is a flowchart showing the processing flow of the obstacle detection processing unit in Embodiment 4. [Modes for carrying out the invention]

[0011] The following describes an obstacle detection system according to an embodiment, with reference to the drawings. The following embodiment is merely an example, and it is possible to combine embodiments as appropriate and modify each embodiment as appropriate. In the drawings, similar components are denoted by the same reference numerals.

[0012] Embodiment 1. The obstacle detection system 1000 in Embodiment 1 will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the obstacle detection system 1000 of Embodiment 1. The obstacle detection system 1000 includes an obstacle detection device 100, a plurality of home doors 200, a plurality of imaging devices 300, and a comprehensive control unit 400. Hereinafter, when not distinguishing between the plurality of home doors 200, they are simply referred to as the home door 200, and when not distinguishing between the plurality of imaging devices 300, they are simply referred to as the imaging device 300. Further, the obstacle detection device 100 includes an image analysis server 101 and an obstacle detection controller 104. The image analysis server 101 includes an image acquisition unit 102 and an obstacle detection processing unit 103.

[0013] The imaging device 300 is provided corresponding to each of the home doors 200 and images the image around the home door 200 including the home door 200. That is, the imaging device 300 is arranged for each monitoring area around the home door 200 and images the image around the home door 200. The imaging device 300 is, for example, a network camera or the like. FIG. 2 is a schematic diagram showing the home door 200 of Embodiment 1. In FIG. 2, the hatched area is the imaging area 301 of the imaging device 300. Note that the imaging device 300 is not shown in FIG. 2.

[0014] As shown in FIG. 2, the imaging area 301 of the imaging device 300 is set around the home door 200 so as to include the home door 200. Therefore, the imaging area 301 may include homes around the home door 200, lines, and the like.

[0015] In FIG. 1, an individual control panel 201 is provided for each home door 200. Based on the output from the integrated control unit 400, the individual control panel 201 performs various controls related to the door, such as opening and closing control of the door. That is, the individual control panel 201 performs opening and closing control of the home door 200. Also, as shown in FIG. 2, a notification device 202 is provided for each of the home doors 200. The notification device 202 notifies users around the home door 200, by means of the buzzer sounding and the lamp lighting, that there is an obstacle 500 around the home door 200. Here, the user is a person who uses the railway vehicle. Also, the obstacle 500 is something that may obstruct the safe operation of the railway vehicle and the safe operation of the home door 200 around the home door 200. For example, it is a user around the home door 200, a user's carry-on item such as a bag, and a user's cane, etc.

[0016] As shown in FIG. 1, the image acquisition unit 102 is connected to each of the plurality of imaging devices 300 and acquires the image captured by the imaging device 300 from the imaging device 300. That is, the obstacle detection device 100 is provided outside the imaging device 300 and is connected to the imaging device 300. The image acquisition unit 102 outputs the acquired image to the obstacle detection processing unit 103. Based on the input image, the obstacle detection processing unit 103 detects the presence or absence of an obstacle 500 around the home door 200 and generates obstacle information indicating at least the presence or absence of the obstacle 500 around the home door 200. At this time, the obstacle detection processing unit 103 detects the presence or absence of the obstacle 500 in the detection area 302 set around the home door 200 so as to include the home door 200 among the images of the imaging area 301. The obstacle detection processing unit 103 outputs the generated obstacle information to the obstacle detection controller 104. The obstacle detection controller 104 outputs the obstacle information to the individual control panel 201 of the home door 200. Note that the obstacle information may include information indicating the type of the obstacle 500 in addition to the presence or absence of the obstacle 500.

[0017] Since the obstacle detection processing unit 103 is located in the image analysis server 101, which acts as a central server, rather than in the imaging device 300, it can process a larger amount of data compared to a configuration with similar functionality in the imaging device 300, and can perform obstacle detection with higher accuracy than the imaging device 300 alone.

[0018] In Figure 1, the integrated control unit 400 comprises a display unit 401, an opening / closing signal acquisition unit 402, a communication unit 403, and a platform door controller 404. The integrated control unit 400 controls the opening and closing of each platform door 200 by controlling the individual control panels 201. The location of the integrated control unit 400 is not particularly limited; for example, it may be installed in an equipment room on the platform, or it may be installed on the platform in parallel with the platform doors 200.

[0019] The individual control panel 201 transmits information regarding the open / closed status of the platform doors 200 installed on it to the open / closed signal acquisition unit 402, and the open / closed signal acquisition unit 402 transmits this information to the display unit 401 and the communication unit 403. The display unit 401 displays the open / closed status of each platform door 200. Based on the information regarding the open / closed status of the platform doors 200 displayed on the display unit 401, the train driver confirms that all platform doors 200 are closed and then determines that the train can proceed.

[0020] Furthermore, the communication unit 403 receives signals related to the opening and closing control of the platform doors 200, input by station staff, from the station staff control panel 600, and outputs these opening and closing control signals to the platform door controller 404. The platform door controller 404 is connected to the individual control panel 201 of each platform door 200 and outputs the opening and closing control signals to the individual control panel 201.

[0021] Here, signals between the platform doors 200 and the central control unit 400, that is, signals between the individual control panel 201 of each platform door 200 and the open / close signal acquisition unit 402 and platform door controller 404 of the central control unit 400, are transmitted and received based on a communication function compliant with the first protocol in order to establish a communication environment using multiple types of signals. In other words, the individual control panel 201, the open / close signal acquisition unit 402 and platform door controller 404 are connected by a communication line compliant with the first protocol. Furthermore, signals between the imaging device 300 and the obstruction detection device 100, that is, signals between each imaging device 300 and the image acquisition unit 102 of the obstruction detection device 100, are also transmitted and received based on a communication function compliant with the first protocol.

[0022] The first protocol is, for example, the Ethernet method and the RS485 method. When the Ethernet method is used as the first protocol, as described above, a communication environment using multiple types of signals can be established. When the RS485 method is used as the first protocol, a communication environment can be established that allows connection of multiple devices with fewer communication lines.

[0023] The obstruction detection processing unit 103 of the image analysis server 101 is connected to the individual control panel 201 of each platform door 200 via the obstruction detection controller 104. Furthermore, signals between the platform doors 200 and the obstruction detection device 100, that is, signals between the individual control panel 201 of each platform door 200 and the obstruction detection controller 104 of the obstruction detection device 100, are transmitted and received based on a communication function compliant with the second protocol. In other words, the individual control panel 201 and the obstruction detection controller 104 are connected by a communication line compliant with the second protocol. The second protocol transmits and receives data composed of fewer bits compared to the first protocol, resulting in less data transmission and faster communication speed. The second protocol is, for example, I / O communication compliant with the international standard IEC61131-9, i.e., IO-Link. Furthermore, it is not necessary for all communication lines connecting the obstacle detection controller 104 to the individual control panels 201 of each platform door 200 to conform to the second protocol; it is sufficient for some of them to conform to the second protocol.

[0024] For example, if the obstruction detection processing unit 103 detects an obstruction 500 near the platform door 200, it generates information indicating the presence of an obstruction 500 near the platform door 200 as obstruction information and outputs it to the obstruction detection controller 104. The obstruction detection controller 104 outputs this obstruction information to the individual control panel 201 as an I / O signal indicating the presence of an obstruction 500 near the platform door 200.

[0025] Alternatively, for example, if the obstacle detection processing unit 103 does not detect an obstacle 500 around the platform door 200, it may generate information indicating that there is no obstacle 500 around the platform door 200 as obstacle information and output it to the obstacle detection controller 104. The obstacle detection controller 104 outputs this obstacle information to the individual control panel 201 as an I / O signal indicating that there is no obstacle 500 around the platform door 200.

[0026] The individual control panel 201, based on the obstacle information as an I / O signal received from the obstacle detection controller 104, performs various processes such as controlling the opening and closing of the platform doors 200 and notifying that an obstacle 500 is present around the platform doors 200, according to the opening and closing status of the platform doors 200.

[0027] For example, if an I / O signal indicating the presence of an obstruction 500 near the platform door 200 is received during the door closing operation before the door is completely closed, the individual control panel 201 controls the door to stop the closing operation and perform the opening operation. In this case, the individual control panel 201 may also control the notification device 202 to sound a buzzer and light up a lamp, etc.

[0028] Furthermore, for example, if an I / O signal indicating the presence of an obstruction 500 around the platform door 200 is received after the door closing operation is completed, the individual control panel 201 may detect the presence of a user and their belongings remaining between the platform door 200 and the tracks, and control the notification device 202 to sound a buzzer and illuminate a lamp, etc. In this case, the individual control panel 201 may also control the door to open it.

[0029] Furthermore, for example, if an I / O signal is received indicating the presence of an obstruction 500 around the platform door 200 when the door is open, i.e., when a passenger is on board, the individual control panel 201 controls the door to maintain the open state because a passenger attempting to board has been detected. Note that the control of the platform door 200 by the individual control panel 201 is not limited to the above, and an appropriate method may be adopted depending on the opening and closing status of the platform door 200.

[0030] As shown in Figure 1, the communication lines compliant with the second protocol, which connect the obstacle detection controller 104 and the individual control panels 201 of each platform door 200, are connected in a daisy-chain configuration. This reduces the number of wires and simplifies wiring work compared to a system where the obstacle detection controller 104 and each individual control panel 201 of each platform door 200 are directly connected.

[0031] Next, the obstacle detection process performed by the obstacle detection processing unit 103 will be described. Figure 3 is a schematic diagram illustrating the processing of the obstacle detection processing unit 103 in Embodiment 1. Figure 3A is an image of the area around the platform door 200 in a normal state without the obstacle 500, and Figure 3B is an image of the area around the platform door 200 with the obstacle 500 present. The image of the area around the platform door 200 in a normal state without the obstacle 500, as shown in Figure 3A, is saved to the image analysis server 101 or an external cloud, etc., and output to the obstacle detection processing unit. The obstacle detection processing unit 103 compares the image of the area around the platform door 200 in a normal state with the image input from the image acquisition unit 102, and if there is a difference, generates obstacle information indicating the presence of the obstacle 500. In other words, when the obstruction detection processing unit 103 receives an image from the image acquisition unit 102 showing the presence of an obstruction 500 as shown in Figure 3B, it compares it with an image of the area around the platform door 200 under normal conditions, takes the difference, and generates obstruction information indicating the presence of an obstruction 500.

[0032] Furthermore, the obstacle detection processing function performed by the obstacle detection processing unit 103 may be implemented by machine learning. The following describes the case in which the function of the obstacle detection processing unit 103 is implemented by machine learning. Figure 4 is a configuration diagram of the learning device 10 related to the obstacle detection device 100 of Embodiment 1. The learning device 10 includes a data acquisition unit 10a, a model generation unit 10b, and a trained model storage unit 10c.

[0033] The learning device 10 is used to learn the obstacle information output by the obstacle detection processing unit 103. However, it may be connected to the obstacle detection processing unit 103 via a network and may be a separate device from the obstacle detection device 100. Alternatively, the learning device 10 may be built into the obstacle detection device 100. Furthermore, the learning device 10 may reside on a cloud server.

[0034] The data acquisition unit 10a of the learning device 10 acquires images of the area around the platform doors 200 and correct data of obstruction information indicating the presence or absence of obstructions 500 corresponding to those images as learning data.

[0035] The model generation unit 10b learns obstacle information corresponding to images of the platform doors 200 based on training data created from a combination of images of the area around the platform doors 200 output from the data acquisition unit 10a and ground truth data of obstacle information indicating the presence or absence of obstacles 500 corresponding to those images. In other words, the model generation unit 10b generates a learning model that infers optimal obstacle information from images captured by the imaging device 300 and ground truth data of obstacle information indicating the presence or absence of obstacles 500 corresponding to those images. Here, the training data is data that associates images of the area around the platform doors 200 with ground truth data of obstacle information indicating the presence or absence of obstacles 500 corresponding to those images.

[0036] The learning algorithm used by the model generation unit 10b can be any known algorithm, such as supervised learning or unsupervised learning. The trained model storage unit 10c stores the trained model output from the model generation unit 10b.

[0037] Figure 5 is a diagram showing the configuration of the obstacle detection device 100 according to Embodiment 1. The obstacle detection processing unit 103 comprises a data acquisition unit 103a and an inference unit 103b. The data acquisition unit 103a of the obstacle detection processing unit 103 acquires images of the area around the platform door 200 from the image acquisition unit 102. The inference unit 103b infers obstacle information obtained using a trained model. That is, by inputting the images acquired by the data acquisition unit 103a into this trained model, obstacle information inferred from the images can be output.

[0038] Figure 6 is a schematic diagram showing an example of training image data in Embodiment 1. The model generation unit 10b uses image data under various environmental conditions and in the presence of various obstacles 500, as shown in Figure 6, as training data to learn obstacle information corresponding to images around the platform door 200.

[0039] In Figure 6, Figures 6A, 6B, 6C, and 6D are image data of the area around the platform door 200 under normal conditions under various environmental conditions. In addition, in Figure 6, Figures 6E, 6F, 6G, and 6H are image data of the area around the platform door 200 when an obstruction 500 is present under various environmental conditions.

[0040] Figure 6A shows image data when snow and small objects such as insects are present in the detection area 302. Figure 6B shows image data when fog is present in the detection area 302, with the hatched area being the area where fog is present. Figure 6C shows image data when the detection area 302 is illuminated by sunlight, with the hatched area being the area illuminated by sunlight. Figure 6D shows image data when trash is present in the detection area 302. Figure 6E shows image data when a bag, a personal item of the user, is present in the detection area 302 as an obstacle 500. Figure 6F shows image data when a user's cane is present in the detection area 302 as an obstacle 500. Figure 6G shows image data when fog is present in the detection area 302 and a person is present as an obstacle 500, with the hatched area being the area where fog is present. Figure 6H shows image data when the detection area 302 is illuminated by sunlight and a person is present as an obstruction 500, with the hatched area being the area illuminated by sunlight.

[0041] As shown in Figure 6, environmental conditions such as snowfall, fog, and sunlight can cause false detections in obstacle detection. For example, snowfall and fog may be falsely detected as obstacles 500. Strong sunlight may also make obstacle detection difficult. Furthermore, garbage such as plastic bottles and paper waste may also be falsely detected as obstacles 500. Therefore, as shown in Figure 6, the obstacle detection processing unit 103 generates obstacle information using a trained model that has learned image data of the area around the platform door 200 under normal conditions without obstacles 500, under such environmental conditions.

[0042] Furthermore, as shown in Figure 6, the obstacle detection processing unit 103 generates obstacle information using a trained model that has been trained to use image data of the area around the platform door 200 when an obstacle 500 is present, including image data of the user's personal belongings such as a bag and cane, as well as image data of the user under environmental conditions such as fog and sunlight.

[0043] Next, we will describe the hardware configuration examples of each component in Embodiment 1, namely the image acquisition unit 102, obstacle detection processing unit 103, obstacle detection controller 104 of the obstacle detection device 100, the display unit 401, opening / closing signal acquisition unit 402, communication unit 403 of the integrated control unit 400, the platform door controller 404, the conversion device 203 described later, and the fault determination unit 405. Figure 7 is a block diagram showing the hardware configuration examples of each component in Embodiment 1. Each component in Embodiment 1 may be a processing circuit 90 which is dedicated hardware as shown in Figure 7A, or it may be a processor 92 that executes a program stored in memory 94 as shown in Figure 7B.

[0044] As shown in Figure 7A, when each component in Embodiment 1 is dedicated hardware, the processing circuit 90 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-programmable Gate Array), or a combination thereof. Each function of each component in Embodiment 1 may be implemented by the processing circuit 90, or the functions of each part may be implemented together by a single processing circuit 90.

[0045] As shown in Figure 7B, when each configuration in Embodiment 1 is a processor 92, the functions of each part are realized by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in memory 94. The processor 92 realizes each function of each configuration in Embodiment 1 by reading and executing the program stored in memory 94. In other words, each configuration in Embodiment 1 is equipped with memory 94 for storing a program that, when executed by the processor 92, will result in the execution of the steps shown in Figures 8 and 13, which will be described later. These programs can also be said to cause the computer to execute the procedures or methods of each configuration in Embodiment 1.

[0046] Here, processor 92 refers to, for example, a CPU (Central Processing Unit), processing unit, arithmetic unit, processor, microprocessor, microcomputer, or DSP (Digital Signal Processor). Memory 94 may be, for example, a non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable ROM), or EEPROM (Electrically EPROM), or a magnetic disk such as a hard disk or flexible disk, or an optical disk such as a MiniDisc, CD (Compact Disc), or DVD (Digital Versatile Disc).

[0047] Furthermore, for each function of each configuration in Embodiment 1, some may be implemented with dedicated hardware, and some may be implemented with software or firmware. In this way, the processing circuit 90 in Embodiment 1 can implement the above-mentioned functions by hardware, software, firmware, or a combination thereof.

[0048] Next, the processing flow of the obstacle detection system 1000 in Embodiment 1 will be described. Figure 8 is a flowchart of the processing flow of the obstacle detection system 1000 in Embodiment 1.

[0049] In step S101, the imaging device 300 captures an image of the area around the platform door 200, including the platform door 200, and outputs the captured image to the image acquisition unit 102 of the obstruction detection device 100. The image acquisition unit 102 outputs the image to the obstruction detection processing unit 103.

[0050] In step S102, the obstacle detection processing unit 103 detects the presence or absence of an obstacle 500 around the platform door 200 based on the input image and generates obstacle information indicating at least the presence or absence of an obstacle 500 around the platform door 200. The obstacle detection processing unit 103 outputs the generated obstacle information to the obstacle detection controller 104.

[0051] In step S103, the obstacle detection controller 104 outputs obstacle information to the individual control panel 201 of the platform door 200. At this time, the signals between the obstacle detection controller 104 and the individual control panel 201 are transmitted and received based on a communication function compliant with a second protocol, which has a lower communication volume and faster communication speed because it transmits and receives data composed of fewer bits compared to the communication function compliant with the first protocol between the integrated control unit 400 and the individual control panel 201.

[0052] In step S104, the individual control panel 201 of the platform door 200 performs at least one of the following based on the obstruction information: control the opening and closing of the platform door 200, or notify that an obstruction 500 is present around the platform door 200. This concludes the explanation of the processing flow of the obstacle detection system 1000 in Embodiment 1.

[0053] As described above, the obstacle detection device 100 in Embodiment 1 is an obstacle detection device 100 connected to an imaging device 300 that captures an image of the area around the platform door 200 including the platform door 200, and comprises an image acquisition unit 102 that acquires an image from the imaging device 300, an obstacle detection processing unit 103 that has a larger data processing capacity than the imaging device 300, detects the presence or absence of an obstacle 500 around the platform door 200 based on the image, and generates obstacle information that indicates at least the presence or absence of an obstacle 500 around the platform door 200, and an obstacle detection controller 104 that outputs obstacle information to an individual control panel 201 that controls the opening and closing of the platform door 200.

[0054] By adopting the above configuration, the obstacle detection device 100 of Embodiment 1 can process a larger amount of data than the imaging device 300 alone. Therefore, it can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0055] Furthermore, the obstacle detection system 1000 of Embodiment 1 includes an obstacle detection device 100, an individual control panel 201 provided on the platform door 200, and an integrated control unit 400 that controls the platform door 200 by controlling the individual control panel 201. With the above configuration, the obstacle detection system 1000 of Embodiment 1 can process a larger amount of data than the imaging device 300 alone, and therefore can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0056] Furthermore, in Embodiment 1, signals between the integrated control unit 400 and the individual control panel 201 provided on the platform door 200 are transmitted and received based on a communication function compliant with the first protocol, while signals between the obstacle detection controller 104 and the individual control panel 201 provided on the platform door 200 are transmitted and received based on a communication function compliant with the second protocol, which has less data and a faster communication speed compared to the first protocol. With the above configuration, the obstacle detection system 1000 of Embodiment 1 can directly transmit signals from the obstacle detection controller 104 to the individual control panel 201 as I / O data composed of fewer bits. As a result, the obstacle detection system 1000 of Embodiment 1 can shorten the time from when an obstacle 500 is detected until the individual control panel 201 performs either the opening and closing control of the platform door 200 or notifies that an obstacle 500 is present around the platform door 200, compared to when all signals between the obstacle detection controller 104 and the individual control panel 201 are transmitted and received based on a communication function compliant with the first protocol.

[0057] Furthermore, the second protocol of Embodiment 1 is I / O communication. As a result, as described above, the obstacle detection system 1000 of Embodiment 1 can directly transmit signals from the obstacle detection controller 104 to the individual control panel 201 as I / O data composed of fewer bits.

[0058] Furthermore, the obstacle detection processing unit 103 of Embodiment 1 includes a data acquisition unit 103a that acquires images, and an inference unit 103b that outputs obstacle information from the images acquired by the data acquisition unit 103a. With the above configuration, the obstacle detection device 100 of Embodiment 1 can output obstacle information inferred from images by inputting images acquired by the data acquisition unit 103a into a trained model. As a result, the obstacle detection device 100 of Embodiment 1 can reduce false detections of obstacles 500 even under various environmental conditions.

[0059] Furthermore, the individual control panel 201 of Embodiment 1 performs at least one of the following based on the obstruction information: control the opening and closing of the platform doors 200 and provide notification that an obstruction 500 is present around the platform doors 200. With the above configuration, the obstruction detection system 1000 of Embodiment 1 can perform control the opening and closing of the platform doors 200 and provide notification that an obstruction 500 is present around the platform doors 200 depending on the presence or absence of an obstruction 500 around the platform doors 200.

[0060] Furthermore, the integrated control unit 400 may be connected to the obstacle detection controller 104 via a communication line compliant with the first protocol, and may monitor the status of the obstacle detection controller 104. In other words, the integrated control unit 400 and the obstacle detection controller 104 may transmit and receive information such as whether or not the obstacle detection controller 104 is malfunctioning.

[0061] Furthermore, the obstacle detection controller 104 only needs to be installed at the station, and the location of the higher-level image analysis server 101 is not particularly limited. In other words, the image analysis server 101 may be installed in the station office or a management center other than the station, or it may be implemented in the cloud.

[0062] Furthermore, in the daisy-chain configuration described above, each individual control panel 201 may be assigned an identification number. In this case, when the imaging device 300 outputs the image it has captured to the image acquisition unit 102, it also outputs the identification number of the individual control panel 201 of the platform door 200 that it has captured. The obstruction detection processing unit 103 then assigns the identification number to the generated obstruction information and outputs it to each individual control panel 201. As a result, only the individual control panel 201 corresponding to the identification number assigned to the obstruction information indicating that an obstruction 500 has been detected can control the platform door 200 corresponding to that obstruction 500.

[0063] Furthermore, it is not necessary for all connections between the individual control panels 201 and the obstacle detection controllers 104 to be in a daisy-chain configuration; some may be in a daisy-chain configuration, while others may be directly connected.

[0064] Furthermore, it is not necessary to provide one imaging device 300 for each platform door 200; one imaging device 300 may be provided for multiple platform doors 200.

[0065] Furthermore, in the obstruction detection system 1000, a conventional obstruction sensor may be used in conjunction with the imaging device 300. A conventional obstruction sensor is installed on the platform door 200, consists of a flat surface sensor, and detects the presence or absence of an obstruction 500 by transmitting scanning light consisting of infrared light or the like, and receiving and measuring the reflected light from the target object. When the obstruction sensor detects an obstruction 500 around the platform door 200, it outputs the obstruction information to the individual control panel 201.

[0066] In this case, for example, the individual control panel 201 may, when it receives obstacle information from either the obstacle detection controller 104 or the obstacle sensor indicating that an obstacle 500 is present around the platform door 200, perform opening and closing control of the platform door 200 and provide notification that an obstacle 500 is present around the platform door 200. With the above configuration, the obstacle detection system 1000 can improve the detection performance of obstacles 500 by, for example, detecting areas that cannot be detected by conventional obstacle sensors, or by acquiring images of the obstacle 500 on the imaging device 300 side and detecting them.

[0067] Furthermore, in the above case, for example, the individual control panel 201 may, upon receiving obstacle information from both the obstacle detection controller 104 and the obstacle sensor indicating the presence of an obstacle 500 around the platform door 200, perform opening and closing control of the platform door 200 and provide notification that an obstacle 500 is present around the platform door 200. With the above configuration, the obstacle detection system 1000 can avoid false detections, for example, if the conventional obstacle sensor makes a false detection due to rain or snow, and the obstacle detection processing unit 103 does not detect an obstacle 500, it can determine that the obstacle 500 does not exist. In addition, this allows the obstacle detection system 1000 to assist station staff who visually check images to determine the presence or absence of an obstacle 500 by refraining from performing opening and closing control by the individual control panel 201, for example, even if an obstacle 500 has been detected by the conventional obstacle sensor, in cases where a false detection is suspected.

[0068] Embodiment 2. The obstacle detection system 1001 in Embodiment 2 will now be described. The obstacle detection system 1001 in Embodiment 2 differs from the obstacle detection system 1000 in Embodiment 1 in that the platform door 200 is equipped with a converter 203 having terminals corresponding to communication lines compliant with the first protocol and terminals corresponding to communication lines compliant with the second protocol. Components similar to those in Embodiment 1 are denoted by the same reference numerals. A detailed description of components similar to those in Embodiment 1 will be omitted, and the description will mainly focus on components that differ from those in Embodiment 1.

[0069] Figure 9 is a block diagram showing the configuration of the obstacle detection system 1001 of Embodiment 2. In Figure 9, the platform door 200 is equipped with a converter 203 having terminals corresponding to communication lines compliant with the first protocol and terminals corresponding to communication lines compliant with the second protocol. The obstacle detection controller 104 and the individual control panel 201 are connected via the converter 203 by communication lines corresponding to the first protocol and the second protocol. Specifically, the obstacle detection controller 104 and the converter 203 are connected by communication lines corresponding to the first protocol, and the converter 203 and the individual control panel 201 are connected by communication lines corresponding to the second protocol.

[0070] Furthermore, although Figure 9 shows the communication lines connecting the imaging device 300 and the image acquisition unit 102, and the communication lines connecting the obstacle detection controller 104 and the converter 203 separately, the system is not limited to this. That is, these communication lines may be partially shared via a demultiplexer. As a result, the obstacle detection system 1001 of Embodiment 2 can communicate from the obstacle detection controller 104 to the converter 203 using the communication lines compliant with the first protocol that were installed when the imaging device 300 was laid. Therefore, the obstacle detection system 1001 of Embodiment 2 does not require the installation of new communication lines directly connecting the obstacle detection controller 104 and the individual control panel 201 in order to realize a communication function compliant with the second protocol between the obstacle detection controller 104 and the individual control panel 201, thereby simplifying construction and reducing costs.

[0071] Furthermore, as shown in Figure 9, it is not necessary for all platform doors 200 to be equipped with a conversion device 203. Some platform doors 200 may be equipped with a conversion device 203, and the individual control panels 201 of platform doors 200 that are not equipped with a conversion device 203 may be connected in a daisy-chain configuration.

[0072] The obstacle detection device in Embodiment 2 is an obstacle detection device 100 connected to an imaging device 300 that captures images of the area around the platform door 200, including the platform door 200, similar to Embodiment 1. The device comprises an image acquisition unit 102 that acquires images from the imaging device 300, an obstacle detection processing unit 103 that has a larger data processing capacity than the imaging device 300, detects the presence or absence of obstacles 500 around the platform door 200 based on the images, and generates obstacle information that indicates at least the presence or absence of obstacles 500 around the platform door 200, and an obstacle detection controller 104 that outputs obstacle information to an individual control panel 201 that controls the opening and closing of the platform door 200.

[0073] By adopting the above configuration, the obstacle detection device 100 of Embodiment 2 can process a larger amount of data than the imaging device 300 alone. Therefore, it can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0074] Furthermore, the obstacle detection system 1001 of Embodiment 2, like Embodiment 1, includes an obstacle detection device 100, an individual control panel 201 provided on the platform door 200, and a comprehensive control unit 400 that controls the platform door 200 by controlling the individual control panel 201. With the above configuration, the obstacle detection system 1001 of Embodiment 2 can process a larger amount of data than the imaging device 300 alone, and therefore can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0075] Furthermore, the platform door 200 of Embodiment 2 is equipped with a converter 203 having terminals corresponding to communication lines compliant with the first protocol and terminals corresponding to communication lines compliant with the second protocol, which has less data volume and faster communication speed compared to the first protocol. The obstacle detection controller 104 and the individual control panels 201 are connected via the converter 203 by communication lines corresponding to the first protocol and communication lines corresponding to the second protocol. With the above configuration, the obstacle detection system 1001 of Embodiment 2 can transmit at least a portion of the signals between the obstacle detection controller 104 and each individual control panel 201 as I / O data composed of fewer bits. As a result, the obstacle detection system 1001 of Embodiment 2 can shorten the time from when an obstacle 500 is detected until the individual control panel 201 performs either the opening and closing control of the platform door 200 or notifies that an obstacle 500 is present around the platform door 200, compared to when all signals between the obstacle detection controller 104 and the individual control panel 201 are transmitted and received based on a communication function compliant with the first protocol.

[0076] Embodiment 3. The obstacle detection system 1002 in Embodiment 3 will now be described. The obstacle detection system 1002 in Embodiment 3 differs from the obstacle detection system 1000 in Embodiment 1 in that the integrated control unit 400 includes a fault determination unit 405 that determines whether or not the imaging device 300 has malfunctioned. Components similar to those in Embodiment 1 are denoted by the same reference numerals. A detailed explanation of the components similar to those in Embodiment 1 will be omitted, and the differences from Embodiment 1 will be described primarily.

[0077] Figure 10 is a block diagram showing the configuration of the obstacle detection system 1002 according to Embodiment 3. In Figure 10, the integrated control unit 400 includes a fault determination unit 405 that determines whether or not the imaging device 300 has malfunctioned. The fault determination unit 405 is connected to the imaging device 300 by a communication line conforming to the first protocol. The imaging device 300 has a self-diagnosis function that diagnoses whether or not a malfunction has occurred in itself. When the imaging device 300 diagnoses that a malfunction has occurred in itself, it outputs fault information to the fault determination unit 405 indicating that the imaging device 300 has malfunctioned. When the fault determination unit 405 receives the fault information of the imaging device 300, it determines that the imaging device 300 has malfunctioned. In addition, the fault determination unit 405 determines that the imaging device 300 is not malfunctioning unless it has determined that the imaging device 300 has malfunctioned.

[0078] Furthermore, if the imaging device 300 does not have a self-diagnosis function to diagnose whether or not a malfunction has occurred in itself, the fault determination unit 405 may periodically output a response request signal to the imaging device 300 and determine that the imaging device 300 has malfunctioned if there is no response from the imaging device 300. The response request signal is a signal to request a response signal from the imaging device 300, and by receiving this response signal from the imaging device 300, the fault determination unit 405 can confirm that the imaging device 300 is not malfunctioning. Periodically refers to the frequency of requests from the fault determination unit 405 to the imaging device 300, which can be arbitrarily set to once every minute, once every 10 minutes, once every hour, etc.

[0079] If the fault detection unit 405 determines that the imaging device 300 has malfunctioned, it outputs fault information to the display unit 401. The fault detection unit 405 then displays the fault information on the display unit 401, thereby notifying station staff of the malfunction of the imaging device 300.

[0080] The obstacle detection device in Embodiment 3 is an obstacle detection device 100 connected to an imaging device 300 that captures images of the area around the platform door 200, including the platform door 200, similar to Embodiment 1. The device comprises an image acquisition unit 102 that acquires images from the imaging device 300, an obstacle detection processing unit 103 that has a larger data processing capacity than the imaging device 300, detects the presence or absence of obstacles 500 around the platform door 200 based on the images, and generates obstacle information that indicates at least the presence or absence of obstacles 500 around the platform door 200, and an obstacle detection controller 104 that outputs obstacle information to an individual control panel 201 that controls the opening and closing of the platform door 200.

[0081] By adopting the above configuration, the obstacle detection device 100 of Embodiment 3 can process a larger amount of data than the imaging device 300 alone. Therefore, it can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0082] Furthermore, the obstacle detection system 1002 of Embodiment 3, like Embodiment 1, includes an obstacle detection device 100, an individual control panel 201 provided on the platform door 200, and a comprehensive control unit 400 that controls the platform door 200 by controlling the individual control panel 201. With the above configuration, the obstacle detection system 1002 of Embodiment 3 can process a larger amount of data than the imaging device 300 alone, and therefore can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0083] Furthermore, the integrated control unit 400 of Embodiment 3 includes a fault determination unit 405 that determines whether or not the imaging device 300 has malfunctioned. With this configuration, the obstacle detection system 1002 of Embodiment 3 can notify station staff that a malfunction has occurred in the imaging device 300.

[0084] Embodiment 4. The obstacle detection system 1003 in Embodiment 4 will now be described. The obstacle detection system 1003 in Embodiment 4 performs obstacle detection using two modes with different detection sensitivities. That is, the obstacle detection system 1003 in Embodiment 4 differs from the obstacle detection system 1000 in Embodiment 1 in that the obstacle detection processing unit 103 detects the presence or absence of an obstacle 500 around the platform door 200 in a first mode under specific conditions set in advance, and detects the presence or absence of an obstacle 500 in a second mode with at least lower detection sensitivity than the first mode under conditions other than the specific conditions. Components similar to those in Embodiment 1 are denoted by the same reference numerals. Furthermore, a detailed explanation of components similar to those in Embodiment 1 will be omitted, and the components different from those in Embodiment 1 will be mainly described.

[0085] If the obstacle detection processing unit 103 is always performing obstacle detection with high sensitivity, the number of obstacle detections will increase, leading to frequent opening and closing control of the platform doors 200 and notification by the notification device 202, which may cause delays in the operation of railway vehicles. Therefore, the obstacle detection system 1003 can solve this problem by pre-setting specific situations, detecting the presence or absence of obstacles 500 in the first mode in the specific situations, and detecting the presence or absence of obstacles 500 around the platform doors 200 in the second mode, which has at least lower detection sensitivity than the first mode, in situations other than the specific situations.

[0086] In the following explanation, we will use the case where a railway vehicle is present within the station premises as a specific example. Furthermore, the case where a railway vehicle is present within the station premises will be referred to as "on the tracks," and the case where a railway vehicle is not present will be referred to as "off the tracks." Additionally, Mode 1 is assumed to have higher detection sensitivity, a wider detection range, and a faster detection response speed than Mode 2, but it will be effective if its detection sensitivity is at least higher than that of Mode 2.

[0087] Figure 11 is a block diagram showing the configuration of the obstacle detection system 1003 of Embodiment 4. As shown in Figure 11, the obstacle detection system 1003 of Embodiment 4 includes a track presence sensor 700. The track presence sensor 700 detects whether or not a railway vehicle is present in the premises and outputs track presence information indicating whether or not a railway vehicle is present in the premises to the obstacle detection processing unit 103. The obstacle detection processing unit 103 then detects the presence or absence of an obstacle 500 in a first mode when a railway vehicle is present, and in a second mode when a railway vehicle is not present.

[0088] The detection sensitivity when detecting obstacles around the platform screen doors 200 will be described in detail. Figure 12 is a schematic diagram showing an example of training image data for Embodiment 4. In Figure 12, Figures 12A, 12C, 12E, and 12G are training images for the first mode, and Figures 12B, 12D, 12F, and 12H are training images for the second mode.

[0089] Figure 12A is image data when only the tip of the cane, which is the obstacle 500, is present in the detection area 302. Figure 12B is image data when the entire cane, which is the obstacle 500, is present in the detection area 302. Figures 12C and 12D are the same image data, showing the case where the detection area 302 is illuminated by sunlight and a person is present as the obstacle 500, with the hatched area being the area illuminated by sunlight. Figure 12E is image data when a part of a bag, which is the user's personal belongings, is present in the detection area 302 as the obstacle 500. Figure 12F is image data when the entire bag, which is the user's personal belongings, is present in the detection area 302 as the obstacle 500. Figures 12G and 12H are the same image data, showing the case where fog is present in the detection area 302 and a person is present as the obstacle 500, with the hatched area being the area where fog is present.

[0090] When training for the first mode, the model generation unit 10b learns image data of the area around the platform door 200 as if an obstacle 500 were present, even if only the tip of the cane is present in the detection area 302, as shown in Figure 12A. On the other hand, when training for the second mode, the model generation unit 10b learns image data of the area around the platform door 200 as if an obstacle 500 is present when the entire cane is present in the detection area 302, as shown in Figure 12B. For example, if only the tip of the cane is present in the detection area 302, as shown in Figure 12A, it learns image data of the area around the platform door 200 as if an obstacle 500 is not present. As a result, in the first mode with high detection sensitivity, the obstacle detection processing unit 103 can detect even just the tip of the cane as an obstacle 500. On the other hand, in the second mode with low detection sensitivity, the obstacle detection processing unit 103 can detect an obstacle 500 only if the entire cane is present in the detection area 302.

[0091] Furthermore, when training for the first mode, the model generation unit 10b learns image data of the area around the platform door 200 when an obstacle 500 is present, in order to detect a person as an obstacle 500 even when the detection area 302 is illuminated by sunlight, as shown in Figure 12C. On the other hand, when training for the second mode, the model generation unit 10b learns image data of the area around the platform door 200 when an obstacle 500 is absent, in the case of the detection area 302 being illuminated by sunlight, as shown in Figure 12D. As a result, in the first mode with high detection sensitivity, the obstacle detection processing unit 103 can detect a person as an obstacle 500 even when the detection area 302 is illuminated by sunlight. On the other hand, in the second mode with low detection sensitivity, the obstacle detection processing unit 103 can determine that an obstacle 500 is absent even when an obstacle 500 is actually present, when the detection area 302 is illuminated by sunlight.

[0092] Furthermore, when training for the first mode, the model generation unit 10b trains the image data of the area around the platform door 200 as if an obstacle 500 were present, even if only a part of the bag is in the detection area 302, as shown in Figure 12E. On the other hand, when training for the second mode, the model generation unit 10b trains the image data of the area around the platform door 200 as if an obstacle 500 is present when the entire bag is in the detection area 302, as shown in Figure 12F. For example, if only a part of the bag is in the detection area 302, as shown in Figure 12E, it trains the image data of the area around the platform door 200 as if an obstacle 500 is not present. As a result, in the first mode with high detection sensitivity, the obstacle detection processing unit 103 can detect even only a part of the bag as an obstacle 500. On the other hand, in the second mode with low detection sensitivity, the obstacle detection processing unit 103 can detect an obstacle 500 only when the entire bag is in the detection area 302.

[0093] Furthermore, when training for the first mode, the model generation unit 10b learns image data of the area around the platform door 200 when an obstacle 500 is present, in order to detect a person as an obstacle 500 even when fog is present in the detection area 302, as shown in Figure 12G. On the other hand, when training for the second mode, the model generation unit 10b learns image data of the area around the platform door 200 when an obstacle 500 is not present, in the case of fog present in the detection area 302, as shown in Figure 12H. As a result, in the first mode with high detection sensitivity, the obstacle detection processing unit 103 can detect a person as an obstacle 500 even when fog is present in the detection area 302. On the other hand, in the second mode with low detection sensitivity, the obstacle detection processing unit 103 can determine that an obstacle 500 is not present even when an obstacle 500 is actually present, if fog is present in the detection area 302.

[0094] Through the above learning process, the model generation unit 10b can generate a learned model corresponding to the first mode and the second mode, which has lower detection sensitivity than the first mode. Then, the obstacle detection processing unit 103 generates obstacle information using the respective learned model when either the first mode or the second mode is active.

[0095] Furthermore, the changes to the detection sensitivity are not limited to the generation of the multiple learning models described above. For example, the detection sensitivity can be changed as appropriate by changing the filter of the imaging device 300 and by changing the threshold for the presence or absence of an obstruction 500 in the same learning model.

[0096] The detection range can be changed, for example, by making the range of the part of the image output from the imaging device 300 that is set as the detection area 302 narrower in the second mode than in the first mode.

[0097] Furthermore, the detection response speed can be changed in the second mode compared to the first mode by increasing the output time from the obstacle detection processing unit 103 to the obstacle detection controller 104.

[0098] Next, the processing flow of the obstacle detection processing unit 103 in Embodiment 4 will be described. Figure 13 is a flowchart showing the processing flow of the obstacle detection processing unit 103 in Embodiment 4.

[0099] In step S201, the obstacle detection processing unit 103 determines whether or not a railway vehicle is present in the premises based on the presence information received from the presence sensor 700, which indicates whether or not a railway vehicle is present in the premises. If the obstacle detection processing unit 103 determines that a railway vehicle is present in the premises (step S201: YES), it proceeds to step S202. If the obstacle detection processing unit 103 determines that a railway vehicle is not present in the premises (step S201: NO), it proceeds to step S208.

[0100] In step S202, the obstacle detection processing unit 103 performs obstacle detection in the first mode.

[0101] In step S203, the obstacle detection processing unit 103 determines whether or not the opening operation of the platform door 200 has started. Here, the individual control panel 201 outputs information indicating the opening and closing status of the door to the obstacle detection processing unit 103 via the obstacle detection controller 104. If the opening operation of the platform door 200 has not started (step S203: NO), the obstacle detection processing unit 103 returns to step S202 and continues obstacle detection in the first mode. If the opening operation of the platform door 200 has started (step S203: YES), the obstacle detection processing unit 103 proceeds to step S204.

[0102] In step S204, the obstacle detection processing unit 103 stops obstacle detection to prevent the false detection of passengers on the railway vehicle as obstacles 500. Alternatively, the obstacle detection processing unit 103 may continue obstacle detection and stop the opening / closing control and notification based on obstacle information by the individual control panel 201. Furthermore, the opening and closing of the platform doors 200 may be performed by a crew member such as a train driver operating a crew control panel, or it may be performed based on an opening / closing command signal from the general control unit 400 to the individual control panel 201.

[0103] In step S205, the obstacle detection processing unit 103 determines whether the closing operation of the platform door 200 has started. If the closing operation of the platform door 200 has not started (step S205: NO), the obstacle detection processing unit 103 returns to step S204 and continues to stop obstacle detection. If the closing operation of the platform door 200 has started (step S205: YES), the obstacle detection processing unit 103 proceeds to step S206.

[0104] In step S206, obstacle detection in the first mode is resumed.

[0105] In step S207, the obstacle detection processing unit 103 determines whether or not a railway vehicle is present in the premises based on the presence information received from the presence sensor 700, which indicates whether or not a railway vehicle is present in the premises. If the obstacle detection processing unit 103 determines that a railway vehicle is present in the premises (step S207: YES), it returns to step S206 and continues obstacle detection in the first mode. If the obstacle detection processing unit 103 determines that there is no railway vehicle present in the premises (step S207: NO), it proceeds to step S208.

[0106] In step S208, the obstacle detection processing unit 103 performs obstacle detection in the second mode. This concludes the explanation of the processing flow of the obstacle detection processing unit 103 in Embodiment 4.

[0107] The obstacle detection device in Embodiment 4 is an obstacle detection device 100 connected to an imaging device 300 that captures images of the area around the platform door 200, including the platform door 200, similar to Embodiment 1. The device comprises an image acquisition unit 102 that acquires images from the imaging device 300, an obstacle detection processing unit 103 that has a larger data processing capacity than the imaging device 300, detects the presence or absence of obstacles 500 around the platform door 200 based on the images, and generates obstacle information that indicates at least the presence or absence of obstacles 500 around the platform door 200, and an obstacle detection controller 104 that outputs obstacle information to an individual control panel 201 that controls the opening and closing of the platform door 200.

[0108] By adopting the above configuration, the obstacle detection device 100 of Embodiment 4 can process a larger amount of data than the imaging device 300 alone. Therefore, it can perform obstacle detection around the platform doors 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0109] Furthermore, the obstacle detection system 1003 of Embodiment 4, like Embodiment 1, includes an obstacle detection device 100, an individual control panel 201 provided on the platform door 200, and a comprehensive control unit 400 that controls the platform door 200 by controlling the individual control panel 201. With this configuration, the obstacle detection system 1003 of Embodiment 4 can process a larger amount of data than the imaging device 300 alone, and therefore can perform obstacle detection around the platform door 200 based on captured images more accurately than when the imaging device 300 alone cannot process a larger amount of data.

[0110] Furthermore, the obstacle detection processing unit 103 of Embodiment 4 detects the presence or absence of obstacles 500 around the platform door 200 in a first mode under specific conditions that have been set in advance, and detects the presence or absence of obstacles 500 around the platform door 200 in a second mode with at least lower detection sensitivity than the first mode under conditions other than those that have been set in advance. With the above configuration, for example, if the presence of a railway vehicle within the station premises is set as a specific condition, the obstacle detection device 100 of Embodiment 4 can change its detection sensitivity depending on whether the vehicle is present or not. As a result, when the vehicle is present, it can detect obstacles 500 without fail using the first mode with high detection sensitivity to enhance stability, and when the vehicle is not present, it can detect obstacles using the second mode with lower detection sensitivity than the first mode. As a result, the obstacle detection processing unit 103 can appropriately detect obstacles when the train is present and when it is not. In particular, the detection sensitivity becomes excessively high when the train is not present, which increases the number of obstacle detections. This reduces the impact of delays in train operations caused by frequent opening and closing control of the platform doors 200 and notification by the notification device 202.

[0111] While the specific conditions mentioned above are those of being on the tracks, the Obstacle Detection Processing Unit 103 is not limited to these. For example, in addition to being on the tracks, the Obstacle Detection Processing Unit 103 may also set season and time of day as specific conditions. In this case, for example, the Obstacle Detection Processing Unit 103 may set the specific conditions as being on the tracks and the season or time of day when the level of congestion on the platform is expected to be normal, and perform obstacle detection in the first mode. Furthermore, even when being on the tracks, in seasons or time of day when congestion on the platform is expected, the Obstacle Detection Processing Unit 103 may perform obstacle detection in a second mode as a condition other than the specific conditions in order to prevent false detections. With the above configuration, the Obstacle Detection Processing Unit 103 can reduce the possibility of false detections due to excessively high detection sensitivity in seasons or time of day when congestion on the platform is expected.

[0112] Furthermore, although it has been explained that the train occupancy sensor 700 outputs train occupancy information to the obstruction detection processing unit 103, station staff or train crew may manually input train occupancy information to the obstruction detection processing unit 103 using the station staff control panel 600 or the train crew control panel.

[0113] Furthermore, while the obstacle detection system 1003 in Embodiment 4 performs obstacle detection using two modes with different detection sensitivities, it is not limited to this. That is, the obstacle detection system 1003 in Embodiment 4 only needs to be able to detect the presence or absence of the obstacle 500 in the first mode in a predetermined specific situation, and in a second mode with at least lower detection sensitivity than the first mode in situations other than the specific situation, and the number of modes may be three or more.

[0114] (Note 1) An obstruction detection device connected to an imaging device that captures images of the area around the platform doors, including the platform doors, An image acquisition unit that acquires the aforementioned image from the imaging device, An obstacle detection processing unit has a larger data processing capacity than the aforementioned imaging device, detects the presence or absence of obstacles around the platform door based on the image, and generates obstacle information indicating at least the presence or absence of obstacles around the platform door; An obstacle detection controller that outputs the obstacle information to an individual control panel that controls the opening and closing of the platform doors, An obstacle detection device equipped with the following features. (Note 2) Obstacle detection device as described in Appendix 1, The individual control panel provided in the platform door, A comprehensive control unit that controls the platform doors by controlling the individual control panels, An obstacle detection system equipped with the following features. (Note 3) The signals between the integrated control unit and the individual control panels are transmitted and received based on a communication function compliant with the first protocol. The signals between the obstacle detection controller and the individual control panel are transmitted and received based on a communication function compliant with a second protocol, which has less data volume and a faster communication speed compared to the first protocol. Obstacle detection system as described in Appendix 2. (Note 4) The aforementioned second protocol is I / O communication. Obstacle detection system as described in Appendix 3. (Note 5) The platform door is equipped with a converter having terminals corresponding to communication lines conforming to a first protocol, and terminals corresponding to communication lines conforming to a second protocol which has less data volume and a faster communication speed compared to the first protocol. The obstacle detection controller and the individual control panel are connected via the converter by a communication line corresponding to the first protocol and a communication line conforming to the second protocol. An obstacle detection system as described in any one of the items from Appendix 2 to Appendix 4. (Note 6) The obstacle detection processing unit comprises a data acquisition unit that acquires the image and an inference unit that outputs the obstacle information from the image acquired by the data acquisition unit. Obstacle detection device as described in Appendix 1. (Note 7) The integrated control unit includes a fault determination unit that determines whether or not the imaging device has malfunctioned. An obstacle detection system as described in any one of the items from Appendix 2 to Appendix 5. (Note 8) The obstruction detection processing unit detects the presence or absence of the obstruction around the platform door in a first mode under specific conditions set in advance, and detects the presence or absence of the obstruction around the platform door in a second mode with at least lower detection sensitivity than the first mode under conditions other than the specific conditions. An obstacle detection device as described in either Appendix 1 or Appendix 6. (Note 9) The system further includes an obstacle sensor that transmits scanning light and receives reflected light from an object to determine the presence or absence of the aforementioned obstacle. An obstacle detection system as described in any one of the following items: Appendix 2 to Appendix 5, and Appendix 7. (Note 10) Based on the obstruction information, the individual control panel performs at least one of the following: control the opening and closing of the platform doors and notify that the obstruction is present around the platform doors. An obstacle detection system as described in any one of the following items: Appendix 2 to Appendix 5, Appendix 7, and Appendix 9. [Explanation of symbols]

[0115] 1000, 1001, 1002, 1003 Obstacle detection system, 100 Obstacle detection device, 101 Image analysis server, 102 Image acquisition unit, 103 Obstacle detection processing unit, 103a Data acquisition unit, 103b Inference unit, 104 Obstacle detection controller, 200 Platform door, 201 Individual control panel, 202 Notification device, 203 Conversion device, 300 Imaging device, 301 Imaging area, 302 Detection area, 400 Integrated control unit, 401 Display unit, 402 Open / close signal acquisition unit, 403 Communication unit, 404 Platform door controller, 405 Fault judgment unit, 500 Obstacle, 600 Station staff operation panel, 700 Occupancy sensor, 10 Learning device, 10a Data acquisition unit, 10b Model generation unit, 10c Learned model storage unit, 90 Processing circuit, 92 processors, 94 memory

Claims

1. An obstruction detection device connected to an imaging device that captures images of the area around the platform doors, including the platform doors, An image acquisition unit that acquires the aforementioned image from the imaging device, An obstacle detection processing unit has a larger data processing capacity than the aforementioned imaging device, detects the presence or absence of obstacles around the platform door based on the image, and generates obstacle information indicating at least the presence or absence of obstacles around the platform door; An obstacle detection controller that outputs the obstacle information to an individual control panel that controls the opening and closing of the platform doors, An obstacle detection device equipped with the following features.

2. Obstacle detection device according to claim 1, The individual control panel provided in the platform door, A comprehensive control unit that controls the platform doors by controlling the individual control panels, An obstacle detection system equipped with the following features.

3. The signals between the integrated control unit and the individual control panels are transmitted and received based on a communication function compliant with the first protocol. The signals between the obstacle detection controller and the individual control panel are transmitted and received based on a communication function compliant with a second protocol, which has less data volume and a faster communication speed compared to the first protocol. Obstacle detection system according to claim 2.

4. The second protocol is I / O communication. The obstacle detection system according to claim 3.

5. The platform door is equipped with a converter having terminals corresponding to communication lines conforming to a first protocol, and terminals corresponding to communication lines conforming to a second protocol which has less data volume and a faster communication speed compared to the first protocol. The obstacle detection controller and the individual control panel are connected via the converter by a communication line corresponding to the first protocol and a communication line conforming to the second protocol. Obstacle detection system according to claim 2.

6. The obstacle detection processing unit comprises a data acquisition unit that acquires the image and an inference unit that outputs the obstacle information from the image acquired by the data acquisition unit. Obstacle detection device according to claim 1.

7. The integrated control unit includes a fault determination unit that determines whether or not the imaging device has malfunctioned. Obstacle detection system according to claim 2.

8. The obstruction detection processing unit detects the presence or absence of the obstruction around the platform door in a first mode under specific conditions set in advance, and detects the presence or absence of the obstruction around the platform door in a second mode with at least lower detection sensitivity than the first mode under conditions other than the specific conditions. Obstacle detection device according to claim 1.

9. The system further includes an obstacle sensor that transmits scanning light and receives reflected light from an object to determine the presence or absence of the aforementioned obstacle. Obstacle detection system according to claim 2.

10. Based on the obstruction information, the individual control panel performs at least one of the following: control the opening and closing of the platform doors and notify that the obstruction is present around the platform doors. Obstacle detection system according to claim 2.

Images