Detection system and method

Abstract

To detect the occurrence of obstacles around railway tracks without fail. [Solution] The detection system 1 is mounted on a railway vehicle traveling on rail R and comprises an imaging unit 21 that outputs an image of the rail R, and a detection unit 43 that inputs the image into a trained model and detects the presence or absence of an obstacle OB around the rail R. The trained model is a machine learning model trained using embedded images generated by embedding an image of an obstacle randomly selected from a predetermined number of obstacles in a randomly determined position around the rail R in the image output by the imaging unit 21 as training data.

Description

[Technical Field]

[0001] The present disclosure relates to detection systems and methods. [Background technology]

[0002] Patent document 1 describes a railway track abnormality detection device that includes an image reading means for reading an image of the railway track, a storage means for storing the read image, a comparison means for comparing the inspection image read by the image reading means with a reference image previously stored in the storage means, and an output means for identifying and outputting abnormal locations on the railway track based on the comparison results. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 3-250103 Summary of the Invention [Problem to be solved by the invention]

[0004] The anomaly detection device described above identifies anomalies in the railway track by comparing the inspection image with a reference image stored in advance in a storage means. However, various obstacles are expected to occur around the railway track. Since there are various types of obstacles and they exist in various positions, it is desirable to detect all of the obstacles without missing any.

[0005] The present disclosure aims to provide a detection system and a detection method that can detect the occurrence of obstacles around railway tracks without fail. [Means for solving the problem]

[0006] The detection system of the present disclosure is [1] "a detection system comprising: an imaging unit mounted on a railway vehicle traveling on rails, which outputs an image of the rail; and a detection unit which inputs the image into a trained model and detects the presence or absence of an obstacle around the rail, wherein the trained model is a machine learning model trained using image data as training data generated by embedding an image of an obstacle randomly selected from a predetermined number of obstacles in the image output by the imaging unit at a randomly determined position around the rail in the image."

[0007] The detection method of the present disclosure is [6] "a detection method executed by a detection system, comprising: an imaging step of outputting an image of the rail using an imaging unit mounted on a railway vehicle traveling on the rail; and a detection step of inputting the image into a trained model to detect the presence or absence of an obstacle around the rail, wherein the trained model is a machine learning model trained using image data as training data generated by embedding an image of an obstacle randomly selected from a predetermined number of obstacles in the image output by the imaging unit at a randomly determined position around the rail in the image."

[0008] In this detection system or method, an imaging unit mounted on a railway vehicle traveling on rails can efficiently capture images of the area around the rails. The captured images are then input into a trained model to detect the presence or absence of obstacles around the rails. The trained model used here is trained using images of obstacles randomly selected from multiple obstacles in the captured images embedded at random positions around the rails as training data, and is therefore constructed as a model that can detect situations in which multiple types of obstacles exist in various positions around the rails. This makes it possible to detect all obstacles around the rails.

[0009] The detection system of the present disclosure may be [2] "the detection system described in [1] above, further comprising a learning unit that uses the training data to train the machine learning model and construct the trained model." The detection method of the present disclosure may be [7] "the detection method described in [6] above, further comprising a learning step that uses the training data to train the machine learning model and construct the trained model." In this case, a trained model can be constructed that can detect situations where multiple types of obstacles are present at various positions around the rail.

[0010] The detection system of the present disclosure may be the detection system described in [2] above, further comprising: [3] a storage unit that pre-stores a plurality of combinations of information including information on the type of obstacle and the shape of the obstacle; wherein the learning unit determines an obstacle and a shape of the obstacle by randomly selecting and referring to the plurality of combinations of information stored in the storage unit; randomly determines an embedding position around the rail in the captured image; and generates the training data by embedding an image corresponding to the determined obstacle and shape at the embedding position in the captured image. The detection method of the present disclosure may be the detection method described in [7] above, [8] wherein the learning step determines an obstacle and a shape of the obstacle by randomly selecting a plurality of combinations of information including information on the type of obstacle and the shape of the obstacle; randomly determines an embedding position around the rail in the captured image; and generates the training data by embedding an image corresponding to the determined obstacle and shape at the embedding position in the captured image. In this case, if combinations of obstacle types and their shapes are set in advance as realistic combinations, it is possible to generate images in which images of obstacles that could actually exist are embedded as training data. In addition, by randomly determining the embedding positions of the obstacle images in the images based on the position of the rail, it is possible to generate images in which images of obstacles are embedded in positions in which they could actually exist as training data. As a result, it is possible to efficiently build a trained model that can detect situations in which multiple types of obstacles exist in various positions around the rail.

[0011] The detection system of the present disclosure may be [4] "the detection system described in [3] above, in which the learning unit generates the training data using the generation AI by specifying the embedding position and inputting a prompt indicating the determined obstacle and the shape to the generation AI." The detection method of the present disclosure may be [9] "the detection method described in [8] above, in which the learning step generates the training data using the generation AI by specifying the embedding position and inputting a prompt indicating the determined obstacle and the shape to the generation AI." In this case, by using the generation AI, it is not necessary to prepare multiple obstacle images in advance, and the embedding position of the obstacle image can be easily specified. As a result, it is possible to easily generate images in which images of obstacles that may actually exist are embedded at various embedding positions as training data.

[0012] The detection system of the present disclosure may be [5] "the detection system according to any one of [1] to [4] above, wherein the imaging unit has a downward-facing camera that captures an image diagonally downward relative to the horizontal direction, and the detection unit detects the presence or absence of the obstacle based on the image captured by the downward-facing camera." For example, when the imaging unit captures an image of the area directly ahead, the area around the rail in the captured image may appear at the bottom of the image, making it difficult to accurately determine the situation around the rail. However, by using an image captured by a downward-facing camera, it is possible to obtain an image capturing the area around the rail from the front, and by making a determination based on that image, it is possible to more accurately determine the presence or absence of an obstacle. [Effects of the Invention]

[0013] According to the present disclosure, it is possible to detect the occurrence of obstacles around railway tracks without fail. [Brief explanation of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram showing the configuration of a detection system. [Figure 2] FIG. 2 is a block diagram showing the configuration of the imaging device. [Figure 3] FIG. 3 is a block diagram showing a general hardware configuration of a processing server. [Figure 4] FIG. 4 is a block diagram showing the configuration of the processing server. [Figure 5] FIG. 5 is a diagram schematically illustrating the positional relationship between the imaging device and the target object in the embodiment. [Figure 6] FIG. 6 is a flowchart illustrating an example of a detection method in the detection system. [Figure 7] FIG. 7 is a diagram showing an example of data of an obstacle list stored in the processing server. [Figure 8] FIG. 8 is a diagram showing an example of a captured image to be processed by the detection system. [Figure 9] FIG. 9 is a diagram showing an example of an embedded image that is a processing result of the detection system. [Figure 10] FIG. 10 is a flowchart showing an example of a method for learning a trained model. DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, preferred embodiments of the detection system according to the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same or corresponding parts are designated by the same reference numerals, and duplicated explanations will be omitted.

[0016] [Detection system configuration]

[0017] First, the overall configuration of a detection system 1 according to an embodiment will be described with reference to FIG. 1. FIG. 1 is a diagram showing the configuration of the detection system 1. The detection system 1 is a system that detects the occurrence of an obstacle around the rails of a railway track. The railway track includes, for example, a pair of rails R and sleepers T (see FIG. 5) installed on a trackbed. An obstacle OB around the rail R is, for example, an object such as a can or trash that may hinder the travel of a railway vehicle. In this example, the detection system 1 performs the above detection from an image capturing the area around the rail R, including the rail R.

[0018] The detection system 1 includes an imaging device 2, an intermediate server 3, a processing server 4, a database 5, and a display device 6. The imaging device 2 is connected to the intermediate server 3 via a network N. The intermediate server 3 is connected to the processing server 4 via the network N. Images from the imaging device 2 are transmitted to the processing server 4 via the intermediate server 3. The number of intermediate servers 3 may be any number equal to or greater than one. The processing server 4 processes images from the imaging device 2. The processing server 4 is connected to a database 5 and a display device 6 via the network N. Images from the imaging device 2 may be transmitted to the processing server 4 without passing through the intermediate server 3. Alternatively, images from the imaging device 2 may be transmitted to the database 5 without passing through both the intermediate server 3 and the processing server 4. In this case, the processing server 4 acquires images from the database 5 and processes the images. The database 5 stores images processed by the processing server 4. The display device 6 outputs the processing results of the processing server 4. The processing server 4 may be connected to the database 5 and the display device 6 via a local cable. The imaging device 2 is mounted on a railway vehicle. On the other hand, the intermediate server 3, processing server 4, database 5, and display device 6 do not have to be mounted on a railway vehicle, but may be placed, for example, in facilities (such as stations) along the railway line.

[0019] The imaging device 2 captures images of the area around the rail R, including the rail R. A railway vehicle is a vehicle that runs on the rail R, such as a locomotive or a train. The imaging device 2 is mounted, for example, on the lead car of the railway vehicle. As shown in FIG. 2, the imaging device 2 includes an imaging unit 21, a computer 22, a sensor 23, a communication device 24, an input device 25, a display device 26, and a power supply 27. FIG. 2 is a block diagram showing the configuration of the imaging device 2. A user inputs operating conditions to the above devices into the input device 25. The display device 26 outputs images captured by the imaging unit 21. The power supply 27 supplies power to the devices that make up the imaging device 2.

[0020] The imaging unit 21 includes an upward-facing camera 21a, a front-facing camera 21b, and a downward-facing camera 21c. The cameras 21a to 21c are, for example, monocular cameras, stereo cameras, or TOF (Time Of Flight) cameras. When the cameras 21a to 21c are monocular cameras, the portability of the imaging device 2 and the ease of installing the imaging device 2 on a railway vehicle are improved. Each of the cameras 21a to 21c includes an imaging lens 21d, a control unit 21e, and a status display unit 21f.

[0021] Upward-facing camera 21a captures an image diagonally upward with respect to the horizontal direction. Upward-facing camera 21a is mounted on the railway vehicle so that imaging lens 21d faces diagonally upward with respect to the horizontal direction. Front-facing camera 21b captures an image directly ahead in the direction of travel of the railway vehicle. Front-facing camera 21b is mounted on the railway vehicle so that imaging lens 21d faces horizontally. Downward-facing camera 21c captures an image diagonally downward with respect to the horizontal direction. By using cameras 21a to 21c that are appropriate for the position of the object to be imaged, it is possible to preferably image the object.

[0022] The cameras 21a to 21c output a plurality of captured images, including captured images of the rail R and its surroundings, to the calculator 22. The plurality of captured images may include captured images that do not capture the rail R, in addition to captured images that capture the rail R and its surroundings. The cameras 21a to 21c may output only captured images that capture the rail R, without outputting captured images that do not capture the rail R. Each of the plurality of captured images may be a video or a still image. The frame rate of each of the cameras 21a to 21c is, for example, 5 fps to 60 fps, and is, for example, 10 fps. The shutter speed of each of the cameras 21a to 21c is, for example, 1 ms to 5 ms, and is, for example, 2 ms. The frame rate and shutter speed may be appropriately adjusted by the control unit 21e depending on the traveling speed of the railway vehicle. The imaging unit 21 may include at least one of the cameras 21a to 21c.

[0023] The computer 22 is connected to the imaging unit 21, the sensor 23, the communication device 24, the input device 25, and the display device 26, and controls these devices. The computer 22 also stores the captured images from the imaging unit 21. The sensor 23 acquires position information or acceleration information of the railway vehicle and outputs it to the computer 22. The sensor 23 is, for example, a GPS sensor or an acceleration sensor. The computer 22 transmits the position information or acceleration information of the railway vehicle in addition to the captured images to the intermediate server 3 via the communication device 24.

[0024] 3 is a block diagram showing a general hardware configuration of the processing server 4. The processing server 4 includes a CPU (processor) 101 that executes an operating system, application programs, etc., a main memory unit 102 consisting of ROM and RAM, an auxiliary memory unit 103 consisting of a hard disk, flash memory, etc., a communication control unit 104 consisting of a network card or wireless communication module, input devices 105 such as a keyboard and a mouse, and output devices 106 such as a display and a printer.

[0025] Each functional element of the processing server 4, which will be described later, is realized by loading predetermined software onto the CPU 101 or main memory unit 102, operating the communication control unit 104, input device 105, output device 106, display device 6, etc. under the control of the CPU 101, and reading and writing data from and to the main memory unit 102 or auxiliary memory unit 103. Data and databases required for processing are stored in the main memory unit 102 or auxiliary memory unit 103.

[0026] When a moving image is transmitted from the imaging device 2, the processing server 4 divides the moving image into frames to generate still images. The processing server 4 adds position information and acceleration information of the railway vehicle to the still images, and stores the still images and the information in a predetermined storage device. The processing server 4 performs image processing on the generated still images.

[0027] 4 is a block diagram showing the configuration of the processing server 4. The processing server 4 includes a learning unit 41, a generating unit 42, a detecting unit 43, and a table storage unit 44 as functional components.

[0028] The learning unit 41 trains a learning model by machine learning using teacher data (training data) to learn (construct) a learned model. The learning model is, for example, a machine learning model such as a convolutional neural network (CNN). The generation unit 42 converts the number of pixels of the captured image to generate a converted image that is to be detected by the detection unit 43. The detection unit 43 inputs the converted image to the learned model and detects the presence or absence of an obstacle around the rail R based on the output of a weighting calculation for the converted image in the learned model.

[0029] [Detection method]

[0030] Next, a detection method according to an embodiment using the detection system 1 will be described. Fig. 5 is a diagram schematically illustrating the positional relationship between the imaging device 2 and a pair of rails R. In this example, the imaging device 2 uses a downward-facing camera 21c as the imaging unit 21. The downward-facing camera 21c captures images at an angle of 45 degrees diagonally downward relative to the horizontal, for example.

[0031] FIG. 6 is a flowchart showing an example of a detection method in the detection system 1 as flow S1.

[0032] In step S11 (imaging step), the imaging unit 21 outputs a captured image showing the rail R and its surroundings. The imaging device 2 transmits the captured image to the processing server 4. For example, the captured image shows the rail R, sleepers T, and obstacles OB (see FIG. 5) located around the rail R. The number of obstacles OB around the rail R can be any number equal to or greater than one.

[0033] In step S12, the generation unit 42 generates a converted image by performing conversion such as conversion of the number of pixels on the captured image. This conversion is performed so that the image conditions such as the number of pixels in the converted image match the image conditions of the captured image and the embedded image included in the training data used for learning by the learning unit 41, which will be described later.

[0034] In step S13 (detection step), the detection unit 43 inputs the converted image into the trained model to detect the presence or absence of an obstacle OB around the rail R. Specifically, the detection unit 43 inputs the converted image into the trained model, and the trained model performs a weighting calculation on the converted image. The detection unit 43 detects the presence or absence of an obstacle OB based on the output data of the result of the weighting calculation. The trained model is used to detect the presence or absence of an obstacle OB.

[0035] In step S14, the detection unit 43 outputs the detection result. The detection unit 43 may display the processing result on the display device 6, may store the processing result in a predetermined storage device such as a memory or a database, or may transmit the processing result to another computer system.

[0036] [Learning Method]

[0037] Next, an example of a training method for a trained model used in the detection method according to the above embodiment will be described. Fig. 7 shows an example of data of an obstacle list stored in the table storage unit 44 used in the training method, Fig. 8 shows an example of a captured image to be processed by the training method, and Fig. 9 shows an example of an embedded image that is the processing result of the detection method. Fig. 10 is a flowchart showing an example of a training method for a trained model as flow S2. Flow S2 corresponds to a learning step.

[0038] In step S21, the learning unit 41 acquires captured images showing the rail R and its surroundings from the database 5. These captured images generally do not show the obstacle OB. Then, the learning unit 41 pads the captured images by adjusting the brightness (luminance) of the captured images in various ways (step S22). Then, the following processes of steps S23 to S26 are executed on the captured images.

[0039] In step S23, an embedding range A1 (see FIG. 8) including the rail R and its surroundings in the captured image G1 to be processed is set. The embedding range A1 is set in response to an input from the user via an input device 105, such as a mouse, of the processing server 4. Alternatively, the embedding range A1 may be set by the processing server 4 detecting the position of the rail R in the captured image G1 by segmentation using a trained model different from the trained model to be processed in flow S2, and automatically setting a range including the surroundings of the rail R according to the position.

[0040] Next, in step S24, the learning unit 41 randomly determines the embedding position of the obstacle image around the rail R within the embedding range A1 of the captured image G1 set in step S23. For example, the embedding position is determined using two-dimensional coordinates on the image. After that, in step S25, the learning unit 41 randomly selects and refers to obstacle information corresponding to the obstacle image to be embedded at the embedding position in the captured image G1 from multiple pieces of obstacle information (combined information) included in the obstacle list stored in the table storage unit 44, thereby determining the type, shape, and color of the obstacle to be embedded.

[0041] 7, the obstacle information included in the obstacle list includes the type of obstacle, information about the shape of the obstacle, and information about the color of the obstacle. For example, one piece of obstacle information is composed of text data that lists "dust" indicating the type of obstacle, "box-shaped" indicating the shape of the obstacle, and "white" indicating the color of the obstacle. 7, the learning unit 41 selects, for example, the obstacle information "dust, box-shaped, gray" from among the plurality of pieces of obstacle information included in the obstacle list. As a result, gray box-shaped dust is determined as the obstacle to be embedded at the embedding position in the captured image G1.

[0042] Then, in step S26, the learning unit 41 calls a pre-stored generation AI and inputs to the generation AI the captured image G1 and a prompt instructing the generation AI to embed an obstacle image corresponding to the type, shape, and color included in the selected obstacle information into the captured image G1. An example of the generation AI used in this case is Stable Diffusion. When inputting the prompt to the generation AI, the learning unit 41 also inputs information specifying the embedding position of the obstacle image to the generation AI. At this time, the learning unit 41 preferably includes a Japanese word identifying the obstacle, such as "garbage" (trash), in the input prompt, and also automatically adds an English word, such as "trash." This improves the accuracy of the generated image. As a result of inputting the captured image and the prompt, the learning unit 41 acquires an embedded image in which an obstacle image is embedded at the embedding position of the captured image G1 based on the output of the generation AI. The learning unit 41 repeats the processes of steps S24 to S26 described above for one captured image G1, the number of times the number of embedded images required as training data.

[0043] 9 is a diagram showing examples of embedded images acquired by the learning unit 41 using the processing shown in steps S21 to S26. In this way, the learning unit 41 acquires embedded images G2 to G7 in which images of various types of obstacles OB are embedded at random positions around the rail R. For example, the acquired images include images in which images of the same type of obstacle OB are arranged at different positions as shown in embedded images G2 and G3, images in which images of obstacles OB of different shapes and different colors are embedded as shown in embedded images G4, G5, and G7, and an image in which images of multiple obstacles OB are embedded as shown in embedded image G6.

[0044] Next, in step S27, the learning unit 41 assigns annotation information indicating the presence or absence of obstacles to the multiple embedded images acquired from the generation AI and the multiple captured images acquired in step S22 as training data, and sequentially stores the embedded images and captured images with the annotation information assigned as training data in a predetermined storage device. This annotation information may be information stored in the database 5 or information input by the user.

[0045] In step S28, the learning unit 41 performs training of the learning model using the teacher data stored in the storage device. The learning unit 41 adjusts the parameters of the learning model until a predetermined termination condition is met. The learning unit 41 stores the trained model in a predetermined storage device. This trained model is used in step S13. [Action and effect]

[0046] According to the detection system 1 and detection method described above, the surroundings of the rail R can be efficiently imaged by the imaging device 2 mounted on a railway vehicle traveling on the rail R. Then, the captured image is input into a trained model to detect the presence or absence of an obstacle OB around the rail R. The trained model used here is trained using as training data embedded images in which images of obstacles OB randomly selected from multiple obstacles in the captured image are embedded at random positions around the rail R, and is therefore constructed as a model that can detect situations in which multiple types of obstacles OB are present at various positions around the rail R. This makes it possible to detect the occurrence of obstacles OB around the rail R without omission.

[0047] The detection system 1 of the present disclosure further includes a learning unit 41 that uses teacher data to learn a machine learning model and construct a trained model. In this case, a trained model can be constructed that can detect situations in which multiple types of obstacles OB are present at various positions around the rail R.

[0048] The detection system 1 of the present disclosure further includes a table storage unit 44 that pre-stores multiple pieces of obstacle information, including information on the type and shape of the obstacle. The learning unit 41 randomly selects and references the multiple pieces of obstacle information stored in the table storage unit 44 to determine the obstacle and its shape, randomly determines its embedding position around the rail R in the captured image, and embeds an image corresponding to the determined obstacle and shape at the embedding position in the captured image to generate training data. In this case, by pre-setting combinations of obstacle types and their shapes as realistically conceivable combinations, it is possible to generate images as training data in which images of obstacles that may actually exist are embedded. In addition, by randomly determining the embedding position of the obstacle image in the image based on the position of the rail R, it is possible to generate images as training data in which images of obstacles are embedded at positions where they may actually exist. As a result, it is possible to efficiently build a trained model that can detect situations in which multiple types of obstacles OB are present at various positions around the rail R.

[0049] In the detection system 1 of the present disclosure, the learning unit 41 generates training data using the generation AI by specifying the embedding position and inputting a prompt indicating the determined obstacle and shape to the generation AI. In this case, by using the generation AI, it is not necessary to prepare multiple obstacle images in advance, and the embedding position of the obstacle image can be easily specified. As a result, it is possible to easily generate images in which images of obstacles that may actually exist are embedded at various embedding positions as training data.

[0050] In the detection system 1 of the present disclosure, the imaging device 2 has a downward-facing camera 21c that captures images diagonally downward relative to the horizontal direction, and the detection unit 43 detects the presence or absence of an obstacle OB based on the image captured by the downward-facing camera 21c. For example, when the imaging device 2 captures an image of the area directly ahead, the area around the rail R may appear at the bottom of the image, making it difficult to accurately determine the situation around the rail R. However, by using the image captured by the downward-facing camera 21c, it is possible to obtain an image capturing the area around the rail R from the front, and by making a determination based on that image, it is possible to more accurately determine the presence or absence of an obstacle OB.

[0051] Although the embodiments of the present invention have been described above, the embodiments of the present invention are not limited to the above-described embodiments.

[0052] For example, in the learning method described above, step S22 may be omitted. Also, in the determination method described above, step S12 may be omitted. Also, steps S23 to S25 may be executed in a different order. [Explanation of symbols]

[0053] 1...detection system, 21...imaging unit, 21c...downward-facing camera, 41...learning unit, 43...detection unit, G1...captured image, G2 to G7...embedded images, R...rail, OB...obstacle.

Claims

1. an imaging unit that is mounted on a railway vehicle that runs on a rail and outputs an image of the rail; A detection unit that inputs the captured image into a trained model and detects the presence or absence of an obstacle around the rail; Equipped with The trained model is a machine learning model trained using image data generated by embedding an image of an obstacle randomly selected from a plurality of predetermined obstacles in a randomly determined position around the rail in the captured image output in advance by the imaging unit as training data. Detection system.

2. A learning unit that learns the machine learning model using the training data to construct the trained model. The detection system of claim 1 .

3. A storage unit is provided for storing in advance a plurality of combination information including information on the type of an obstacle and the shape of the obstacle, The learning unit is determining an obstacle and a shape of the obstacle by randomly selecting and referring to the plurality of pieces of combination information stored in the storage unit; Randomly determining an embedding position around the rail in the captured image; generating the training data by embedding an image corresponding to the determined obstacle and shape at the embedding position in the captured image; The detection system of claim 2 .

4. The learning unit generates the training data using the generation AI by specifying the embedding position and inputting a prompt indicating the determined obstacle and the shape to the generation AI. The detection system of claim 3 .

5. the imaging unit has a downward camera that captures an image obliquely downward with respect to a horizontal direction, The detection unit detects the presence or absence of the obstacle based on the captured image captured by the downward camera. A detection system according to any one of claims 1 to 4.

6. A detection method implemented by a detection system, comprising: an imaging step of outputting an image of the rail using an imaging unit mounted on a railway vehicle running on the rail; A detection step of inputting the captured image into a trained model to detect the presence or absence of an obstacle around the rail; Equipped with The trained model is a machine learning model trained using image data generated by embedding an image of an obstacle randomly selected from a plurality of predetermined obstacles in a randomly determined position around the rail in the captured image output in advance by the imaging unit as training data. Detection methods.

7. Further comprising a learning step of learning the machine learning model using the training data to construct the trained model. The detection method according to claim 6.

8. In the learning step, determining an obstacle and a shape of the obstacle by randomly selecting a plurality of combination information including information on the type of the obstacle and the shape of the obstacle; Randomly determining an embedding position around the rail in the captured image; generating the training data by embedding an image corresponding to the determined obstacle and shape at the embedding position in the captured image; The detection method according to claim 7.

9. In the learning step, the embedding position is specified for the generating AI, and a prompt indicating the determined obstacle and the shape is input, whereby the generating AI is used to generate the training data. The detection method according to claim 8.

Images