Determination device, training device, and program

Abstract

A determination device (30) comprises a control unit (31) that, upon receipt of input of at least one captured image including a place where a mirror is disposed on a road, extracts at least one mirror image including an at least partial region of the mirror from the at least one captured image which has been input, and that uses a trained model to determine fogging of the at least partial region of the mirror included in the at least one extracted mirror image.

Description

Determination device, learning device, and program

[0001] The present disclosure relates to a determination device, a learning device, and a program.

[0002] There are known techniques for detecting an object reflected in a convex mirror installed on a road from an image of the convex mirror. For example, Non-Patent Document 1 proposes a method for detecting an object reflected on the mirror surface of a convex mirror using an in-vehicle camera.

[0003] Feng Yuanchao and three others, "Detection of approaching objects reflected in convex mirrors using an in-vehicle camera," Production Research, Vol. 72, No. 2, pp. 201-206, 2020

[0004] With conventional technology, it is difficult to detect images when the mirror surface of a convex mirror is cloudy. To improve safety, it is necessary to quickly detect convex mirrors with cloudy surfaces. However, manual inspection of convex mirrors is costly and inefficient.

[0005] In view of the above circumstances, an object of the present disclosure is to enable efficient determination of mirror fogging.

[0006] In one embodiment, the determination device includes a control unit that, when at least one photographed image including a location on a road where a mirror is installed is input, extracts at least one mirror image including at least a partial area of ​​the mirror from the at least one input photographed image, and determines the fogging of at least a partial area of ​​the mirror included in the extracted at least one mirror image using a trained model.

[0007] A program according to an embodiment causes a computer to function as a determination device according to an embodiment.

[0008] A learning device according to one embodiment includes a control unit that acquires a plurality of mirror images each including at least a partial area of ​​a first mirror extracted from a plurality of captured images each including a location on a road where a first mirror is installed, and fogging information included in the plurality of mirror images indicating fogging of at least a partial area of ​​the first mirror, and that generates a trained model that determines fogging of at least a partial area of ​​the second mirror included in the at least one input mirror image by machine learning using a combination of the acquired plurality of mirror images and the fogging information as training data, when at least one mirror image including at least a partial area of ​​a second mirror is input.

[0009] According to the present disclosure, it is possible to efficiently determine the fogging of a mirror.

[0010] FIG. 1 is a block diagram showing a configuration of a determination system according to an embodiment; FIG. 2 is a flowchart showing an example of operation of the determination system; FIG. 3 is a diagram for explaining extraction of a mirror image; FIG. 4 is a diagram for explaining extraction of a mirror image; FIG. 5 is a diagram for explaining extraction of a mirror image; FIG. 6 is a diagram for explaining extraction of a mirror image; FIG. 7 is a diagram for explaining extraction of a mirror image;

[0011] An embodiment will be described below with reference to the drawings.

[0012] In each drawing, the same or corresponding parts are denoted by the same reference numerals. In the description of this embodiment, the description of the same or corresponding parts will be omitted or simplified as appropriate.

[0013] The configuration of a determination system 10 according to this embodiment will be described with reference to FIG.

[0014] The determination system 10 includes a learning device 20 and a determination device 30 .

[0015] The learning device 20 and the determination device 30 are general-purpose computers such as PCs, server computers such as cloud servers, or dedicated computers. "PC" is an abbreviation for personal computer.

[0016] The learning device 20 and the determination device 30 can communicate with each other via a network 40. The network 40 includes, for example, the Internet, at least one WAN, at least one MAN, or any combination thereof. "WAN" is an abbreviation for wide area network. "MAN" is an abbreviation for metropolitan area network. The network 40 may include at least one wireless network, at least one optical network, or any combination thereof. The wireless network is, for example, an ad hoc network, a cellular network, a wireless LAN, a satellite communication network, or a terrestrial microwave network. "LAN" is an abbreviation for local area network.

[0017] An overview of this embodiment will be described with reference to Figure 1. A learning device 20 acquires multiple mirror images, each including at least a partial region of a first mirror, extracted from multiple captured images, each including a location on a road where a first mirror is installed, and fogging information indicating fogging of at least a partial region of the first mirror included in the multiple mirror images. By machine learning using a combination of the acquired multiple mirror images and the fogging information as training data, a trained model is generated that determines fogging of at least a partial region of the second mirror included in the at least one input mirror image when at least one mirror image including at least a partial region of a second mirror is input. A determination device 30, when at least one captured image including a location on a road where a mirror is installed, extracts at least one mirror image including at least a partial region of the mirror from the at least one input captured image, and determines fogging of at least a partial region of the mirror included in the extracted at least one mirror image using the trained model.

[0018] The mirror is a curved mirror installed at intersections, corners, or entrances and exits of facilities such as parking lots. The mirror has a specular surface, which may be circular, elliptical, polygonal, or any other shape. Mirror fogging can be caused by dirt adhering to the mirror surface or cracks. The captured image may be captured by a camera installed on the road, a camera on a vehicle traveling on the road, or a camera on a terminal device held by a pedestrian. In this embodiment, the learning device 20 generates trained models X1 and X2 through machine learning and transmits the trained models X1 and X2 to the determination device 30. The determination device 30 acquires the trained models X1 and X2 and uses the trained models X1 and X2 to extract a mirror image from the input image and determine the fogging of the mirror contained in the mirror image. According to this embodiment, the fogging of a mirror can be determined based on an image of the mirror without the need for manual inspection of the mirror, thereby enabling efficient determination of mirror fogging.

[0019] The configuration of a learning device 20 according to this embodiment will be described with reference to Fig. 1. The learning device 20 includes a control unit 21, a storage unit 22, a communication unit 23, and an input / output unit 24.

[0020] The control unit 21 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU or GPU, or a dedicated processor specialized for specific processing. "CPU" is an abbreviation for central processing unit. "GPU" is an abbreviation for graphics processing unit. An example of the programmable circuit is an FPGA. "FPGA" is an abbreviation for field-programmable gate array. An example of the dedicated circuit is an ASIC. "ASIC" is an abbreviation for application specific integrated circuit. The control unit 21 controls each part of the learning device 20 and executes processing related to the operation of the learning device 20.

[0021] The storage unit 22 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. The semiconductor memory is, for example, a RAM, a ROM, or a flash memory. "RAM" is an abbreviation for random access memory. "ROM" is an abbreviation for read only memory. RAM is, for example, an SRAM or a DRAM. "SRAM" is an abbreviation for static random access memory. "DRAM" is an abbreviation for dynamic random access memory. ROM is, for example, an EEPROM. "EEPROM" is an abbreviation for electrically erasable programmable read only memory. Flash memory is, for example, an SSD. "SSD" is an abbreviation for solid-state drive. Magnetic memory is, for example, an HDD. "HDD" is an abbreviation for hard disk drive. The storage unit 22 functions, for example, as a main storage device, an auxiliary storage device, or a cache memory. The storage unit 22 stores data used in the operation of the learning device 20, such as teacher data, and data obtained by the operation of the learning device 20.

[0022] The communication unit 23 includes at least one communication module. The communication module is, for example, a module compatible with a wired LAN communication standard such as Ethernet (registered trademark), a wireless LAN communication standard such as IEEE 802.11, or a mobile communication standard such as LTE, 4G, or 5G. "IEEE" is an abbreviation for Institute of Electrical and Electronics Engineers. "LTE" is an abbreviation for Long Term Evolution. "4G" is an abbreviation for 4th generation. "5G" is an abbreviation for 5th generation. The communication unit 23 may communicate with the determination device 30. The communication unit 23 receives data used in the operation of the learning device 20 and transmits data obtained by the operation of the learning device 20.

[0023] The input / output unit 24 includes at least one input / output interface. The input / output interface includes a touch screen integrated with a display. The input / output unit 24 accepts operations to input information used in the operation of the learning device 20 and outputs information obtained by the operation of the learning device 20. The input / output unit 24 may be connected to the learning device 20 as an external input device instead of being provided on the learning device 20. Any connection method, such as a Universal Serial Bus (USB), can be used. The input / output unit 24 may not be integrated, but may instead be composed of separate input and output units. In this case, the input unit includes at least one input interface, such as physical keys or a microphone. The output unit includes at least one output interface, such as a display or a speaker.

[0024] The configuration of a determination device 30 according to this embodiment will be described with reference to Fig. 1. The determination device 30 includes a control unit 31, a storage unit 32, a communication unit 33, and an input / output unit .

[0025] The control unit 31 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU or GPU, or a dedicated processor specialized for a specific process. The programmable circuit is, for example, an FPGA. The control unit 31 controls each part of the determination device 30 and executes processes related to the operation of the determination device 30.

[0026] The storage unit 32 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. The semiconductor memory is, for example, a RAM, a ROM, or a flash memory. The RAM is, for example, an SRAM or a DRAM. The ROM is, for example, an EEPROM. The flash memory is, for example, an SSD. The magnetic memory is, for example, an HDD. The storage unit 32 functions as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage unit 32 stores data used in the operation of the determination device 30, such as captured images and trained models, and data obtained by the operation of the determination device 30.

[0027] The communication unit 33 includes at least one communication module. The communication module is, for example, a module compatible with a wired LAN communication standard such as Ethernet (registered trademark), a wireless LAN communication standard such as IEEE 802.11, or a mobile communication standard such as LTE, 4G standard, or 5G standard. The communication unit 33 may communicate with the learning device 20. The communication unit 33 receives data used in the operation of the determination device 30 and transmits data obtained by the operation of the determination device 30.

[0028] The input / output unit 34 includes at least one input / output interface. The input / output interface includes a touch screen integrated with a display. The input / output unit 34 accepts an operation to input information used in the operation of the determination device 30 and outputs information obtained by the operation of the determination device 30. The input / output unit 34 may be connected to the determination device 30 as an external input device instead of being provided in the determination device 30. Any connection method, such as USB, can be used. The input / output unit 34 may not be integrated, but may be composed of separate input and output units. In this case, the input unit includes at least one input interface, such as physical keys or a microphone. The output unit includes at least one output interface, such as a display or a speaker.

[0029] The functions of the learning device 20 or the determination device 30 are realized by executing a program according to this embodiment on a processor serving as the control unit 21 or the control unit 31. That is, the functions of the learning device 20 or the determination device 30 are realized by software. The program causes a computer to execute the operations of the learning device 20 or the determination device 30, thereby causing the computer to function as the learning device 20 or the determination device 30. That is, the computer functions as the learning device 20 or the determination device 30 by executing the operations of the learning device 20 or the determination device 30 in accordance with the program.

[0030] The program can be stored on a non-transitory computer-readable medium. Examples of the non-transitory computer-readable medium include flash memory, magnetic recording devices, optical disks, magneto-optical recording media, and ROMs. The program can be distributed by selling, transferring, or lending portable media such as SD cards, DVDs, or CD-ROMs that store the program. "SD" is an abbreviation for Secure Digital. "DVD" is an abbreviation for digital versatile disc. "CD-ROM" is an abbreviation for compact disc read only memory. The program can also be distributed by storing it in the storage of a server and transferring it from the server to another computer. The program can also be provided as a program product.

[0031] A computer temporarily stores a program stored on a portable medium or transferred from a server in its main storage device. The computer then reads the program stored in the main storage device with its processor and executes processing in accordance with the read program. The computer may also read the program directly from the portable medium and execute processing in accordance with the program. The computer may also execute processing in accordance with the received program each time a program is transferred from the server to the computer. Processing may also be executed using a so-called ASP-type service that realizes functions simply by issuing execution instructions and obtaining results, without transferring the program from the server to the computer. "ASP" is an abbreviation for application service provider. A program is information used for processing by a computer and includes something equivalent to a program. For example, data that is not a direct instruction to a computer but has properties that define computer processing falls under the category of "something equivalent to a program."

[0032] Some or all of the functions of the learning device 20 or the determination device 30 may be realized by a programmable circuit or a dedicated circuit as the control unit 21 or the control unit 31. In other words, some or all of the functions of the learning device 20 or the determination device 30 may be realized by hardware.

[0033] An example of the operation of the determination system 10 will be described with reference to Fig. 2. Of the operations described below, the operation of the learning device 20 corresponds to a learning method, and the operation of the determination device 30 corresponds to a determination method.

[0034] In S1, the control unit 21 of the learning device 20 acquires a plurality of captured images, each of which includes a location on a road where a first mirror is installed. In this example, the control unit 21 acquires the captured images by reading them from the storage unit 22. However, this is not limited to this, and the control unit 21 may acquire the captured images by receiving them from an external server via the communication unit 23. The control unit 21 may also acquire the captured images by communicating with a vehicle or the like traveling on the road via the communication unit 23 and receiving the captured images directly from the vehicle.

[0035] In S2, the control unit 21 acquires multiple mirror images MI extracted from multiple captured images, each including at least a portion of the first mirror, and fogging information indicating the fogging of at least a portion of the first mirror included in the multiple mirror images MI. Examples of mirror images MI including the first mirror are shown in FIGS. 3A to 3C . As shown in FIG. 3A , the mirror image MI may be an image of a rectangular area including the entire circular mirror surface of the first mirror extracted from captured image A. As shown in FIG. 3B , the mirror image MI may be an image of a rectangular area inscribed in the circular mirror surface of the first mirror extracted from captured image B. As shown in FIG. 3C , the mirror image MI may be an image of individual rectangular areas obtained by dividing an image C of a rectangular area inscribed in the circular mirror surface of the first mirror extracted from captured image B. The fogging information may be binary information indicating whether or not the first mirror is fogging, or may be a value indicating the degree of fogging of the first mirror in stages.

[0036] 3A , the control unit 21 displays the captured image A on the screen via the input / output unit 24. When the control unit 21 receives an operation via the input / output unit 24 to designate a rectangular area including the entire circular mirror surface of the first mirror on the screen on which the captured image A is being displayed, the control unit 21 obtains a mirror image MI by extracting an image of the designated rectangular area from the captured image A. The control unit 21 obtains fogging information by receiving an operation via the input / output unit 24 to input whether or not the circular mirror surface of the first mirror is fogging or the degree of fogging of the entire circular mirror surface of the first mirror.

[0037] In the example shown in Fig. 3B, the control unit 21 displays the captured image B on the screen via the input / output unit 24. When the control unit 21 receives an operation via the input / output unit 24 to designate a rectangular area inscribed in the circular mirror surface of the first mirror on the screen on which the captured image B is being displayed, the control unit 21 obtains a mirror image MI by extracting an image of the designated rectangular area from the captured image B. The control unit 21 receives an operation via the input / output unit 24 to input whether or not the rectangular area inscribed in the circular mirror surface of the first mirror is fogging, or the degree of fogging of the rectangular area inscribed in the circular mirror surface of the first mirror, thereby obtaining fogging information.

[0038] In the example shown in FIG. 3C , the control unit 21 displays the captured image B on the screen via the input / output unit 24. When the control unit 21 receives an operation via the input / output unit 24 to designate a rectangular area inscribed in the circular mirror surface of the first mirror on the screen on which the captured image B is displayed, the control unit 21 extracts an image C of the designated rectangular area from the captured image B. The control unit 21 equally divides the extracted image C. The number of divisions may be a fixed number that is preset, may be dynamically set according to the size of the image C, or may be manually set via the input / output unit 24. The control unit 21 acquires images of the individual rectangular areas obtained by dividing the image C as mirror images MI. In FIG. 3C , the number of divisions is nine, so the control unit 21 acquires each of the nine images as a mirror image MI. However, the number of divisions may be any number of two or more; that is, the control unit 21 may acquire each of any number of images that is two or more as a mirror image MI. The control unit 21 acquires fogging information by accepting, via the input / output unit 24, an operation to input whether or not there is fogging in the corresponding rectangular area for each acquired mirror image MI, or the degree of fogging in the corresponding rectangular area.

[0039] In S3, the control unit 21 performs machine learning using a combination of the acquired photographed image and the extracted mirror image MI as training data to generate a trained model X1 that extracts a mirror image from the photographed image. The trained model X1 may extract a mirror image using any model, such as the YOLO (You Only Look Once) model. Furthermore, the control unit 21 performs machine learning using a combination of the acquired multiple mirror images MI and fogging information as training data. When at least one mirror image including at least a partial area of ​​the second mirror is input, the control unit 21 generates a trained model X2 that determines the fogging of at least a partial area of ​​the second mirror included in the input at least one mirror image. The trained model X2 may determine the fogging using any model, such as the VGG (Visual Geometry Group) 16 model. The control unit 21 may standardize the size of the acquired photographed images as a preprocessing step for machine learning.

[0040] In S4, the control unit 21 transmits the generated trained model X1 and trained model X2 to the determination device 30 via the communication unit 23.

[0041] In S5, the control unit 31 of the determination device 30 acquires the trained model X1 and the trained model X2. In this example, the control unit 31 acquires the trained model X1 and the trained model X2 by receiving the trained model X1 and the trained model X2 from the learning device 20 via the communication unit 33. This is not limiting, and the control unit 31 may acquire the trained model X1 or the trained model X2 by receiving it from an external server via the communication unit 33.

[0042] In S6, the control unit 31 of the determination device 30 accepts input of at least one captured image via the communication unit 33. The captured image is an image that includes a location on a road where a mirror is installed. In this embodiment, the control unit 31 may accept input by receiving the captured image from an external server or a terminal device via the communication unit 33. The control unit 31 may also accept input by reading the captured image from the storage unit 32.

[0043] In S7, the control unit 31 extracts at least one mirror image mi that includes at least a partial area of ​​the mirror from at least one captured image that has been received as input.

[0044] In this example, the control unit 31 uses the trained model X1 acquired in step S5 to extract, from the plurality of captured images received as input in step S6, a plurality of mirror images mi that correspond one-to-one to the plurality of captured images as at least one mirror image mi. Examples of mirror images mi that include a mirror are shown in Figures 4A to 4C.

[0045] In the example shown in Figure 4A, the control unit 31 extracts a mirror image mi, which is an image of a rectangular area, from the captured image D using the trained model X1.

[0046] In the example shown in Figure 4B, the control unit 31 uses the learned model X1 to extract from the captured image E a mirror image mi, which is an image of a rectangular area inscribed in the circular mirror surface of the mirror reflected in the captured image E.

[0047] In the example shown in Figure 4C, the control unit 31 uses the trained model X1 to extract, from image F, mirror image miz, which is obtained by equally dividing a rectangular area inscribed in the circular mirror surface of the mirror captured in the captured image E, from mirror image mia, as mirror image mi. Figure 4C shows a simplified diagram of how the mirror image mi is divided. The number of divisions may be any number of two or more, and the control unit 31 may extract any number of images, two or more, as mirror images mi.

[0048] In S8, the control unit 31 selects a mirror image mi for which fogging is to be determined from the plurality of mirror images mi. Any method may be used for the selection.

[0049] As an example of the selection method, the control unit 31 selects a mirror image mi for which fogging is to be determined based on the sizes of the extracted multiple mirror images mi. In this case, the control unit 31 selects a mirror image mi whose size is equal to or greater than a specified value as a mirror image mi for which fogging is to be determined. The specified size value may be set in advance and stored in the storage unit 32. As a result, if the size of the mirror image mi is too small, such as when the mirror image mi is reflected far away in the captured image, the control unit 31 omits the process of determining fogging for that mirror image mi. In this way, the control unit 31 omits mirror images mi for which sufficient accuracy in determining fogging is not expected. As a result, for example, when there are multiple mirror images mi showing the same mirror, the accuracy of the determination can be improved by using only the selected mirror image mi for determining fogging.

[0050] As an example of a selection method, the control unit 31 selects a mirror image mi to be judged using the trained model X2 based on the aspect ratios of the extracted multiple mirror images mi. In this case, the control unit 31 selects a mirror image mi whose aspect ratio is outside a specified range as a mirror image mi to be judged for fogging. The specified range of aspect ratios may be set in advance and stored in the storage unit 32. As a result, if the mirror image mi is shown as a perspective image in the captured image, the image area of ​​the mirror surface is small, and it is expected that sufficient accuracy in the fogging judgment will not be expected, the control unit 31 omits the fogging judgment process for that mirror image mi. As a result, for example, when there are multiple mirror images mi showing the same mirror, using only the selected mirror image mi for the fogging judgment can improve the accuracy of the judgment.

[0051] As an example of the selection method, the control unit 31 selects a mirror image to be judged using the trained model X2 based on the luminance of the extracted multiple mirror images mi. In this case, the control unit 31 selects a mirror image mi whose pixel luminance is equal to or greater than a specified value as the mirror image mi to be judged for fogging. The specified luminance value may be set in advance and stored in the storage unit 32. As a result, if the mirror image mi is a dark image and it is expected that sufficient accuracy will not be expected in the fogging judgment, the control unit 31 omits the fogging judgment process for that mirror image mi. As a result, for example, when there are multiple mirror images mi that show the same mirror, using only the selected mirror image mi for the fogging judgment makes it possible to improve the accuracy of the judgment.

[0052] The control unit 31 may select the mirror image mi by combining the above-mentioned methods. As described above, the control unit 31 selects the mirror image mi for which fogging is to be determined based on the size, aspect ratio, or brightness of the extracted multiple mirror images mi.

[0053] In S9, the control unit 31 determines the fogging of at least a portion of the mirror area included in at least one of the extracted mirror images mi selected in S8. In this example, the control unit 31 determines the fogging using the trained model X2 acquired in S5. The control unit 31 outputs a binary value indicating the presence or absence of fogging or a value indicating the degree of fogging in stages as the fogging determination result.

[0054] When the mirror image mi is divided, the control unit 31 may determine whether or not each divided mirror image mi is cloudy, and determine the percentage of divided mirror images mi that are determined to be cloudy as the degree of cloudiness. For example, when the mirror image mi shown in image F in FIG. 4C is divided into 20 rectangles, the control unit 31 may determine that five of the divided mirror images mi are cloudy. In this case, the control unit 31 calculates the degree of cloudiness of the mirror image mi to be 25%.

[0055] The method of determining the fogging level by the control unit 31 when the mirror image mi is divided is not limited to this, and for example, the control unit 31 may determine the degree of fogging for each of the divided mirror images mi and determine the average value of the determination results as the degree of fogging for the mirror image mi as a whole. For example, the control unit 31 may determine the degree of fogging for each of the divided mirror images mia to miz shown in image F in Fig. 4C and determine the average value of the determination results as the degree of fogging for the mirror image mi.

[0056] In S10, the control unit 31 outputs information indicating the result of the fogging determination. In this example, the control unit 31 outputs information indicating the presence or absence or degree of fogging via the input / output unit 34. Thereafter, the processing operation of the determination device 30 ends.

[0057] The following additional notes are provided regarding the above-described embodiments.

[0058] (Supplementary Item 1) A determination device comprising a control unit that, when at least one captured image including a location on a road where a mirror is installed, extracts from the at least one input captured image at least one mirror image including at least a partial area of ​​the mirror, and determines fogging of at least a partial area of ​​the mirror included in the extracted at least one mirror image using a trained model. (Supplementary Item 2) The determination device according to Supplementary Item 1, wherein, when a plurality of captured images are input as the at least one captured image, the control unit extracts from the input plurality of captured images a plurality of mirror images that correspond one-to-one to the plurality of captured images as the at least one mirror image, and selects from the plurality of mirror images a mirror image for which fogging is to be determined based on the size, aspect ratio, or brightness of the extracted plurality of mirror images. (Supplementary Item 3) A program that causes a computer to function as the determination device according to Supplementary Item 1 or 2. (Supplementary Item 4) A learning device including a control unit that acquires a plurality of mirror images each including at least a partial area of ​​a first mirror, extracted from a plurality of captured images each including a location on a road where the first mirror is installed, and fogging information that indicates fogging of at least a partial area of ​​the first mirror included in the plurality of mirror images, and that is included in the plurality of mirror images; and that generates a trained model that determines fogging of at least a partial area of ​​the second mirror included in the at least one input mirror image by machine learning using a combination of the acquired plurality of mirror images and the fogging information as training data, when at least one mirror image including at least a partial area of ​​a second mirror is input.

[0059] The present disclosure is not limited to the above-described embodiments. For example, two or more blocks shown in the block diagram may be integrated, or one block may be divided. Two or more steps shown in the flowchart may be executed in parallel or in a different order, instead of being executed in chronological order as described, depending on the processing capabilities of the device executing each step, or as needed. Other modifications are possible within the scope of the present disclosure.

[0060] REFERENCE SIGNS LIST 10 Determination system 20 Learning device 21 Control unit 22 Storage unit 23 Communication unit 24 Input / output unit 30 Determination device 31 Control unit 32 Storage unit 33 Communication unit 34 Input / output unit 40 Network

Claims

1. A determination device having a control unit that, when at least one captured image including a location where a mirror is installed on a road is input, extracts at least one mirror image including at least a partial area of ​​the mirror from the at least one input captured image, and determines the fogging of at least a partial area of ​​the mirror included in the at least one extracted mirror image using a trained model.

2. A determination device as described in claim 1, wherein the control unit, when a plurality of photographed images are input as the at least one photographed image, extracts a plurality of mirror images that correspond one-to-one to the plurality of photographed images from the input plurality of photographed images as the at least one mirror image, and selects from the plurality of mirror images a mirror image for determining the fogging based on the size, aspect ratio, or brightness of the extracted plurality of mirror images.

3. A program for causing a computer to function as the determination device according to claim 1 or 2.

4. A learning device having a control unit that acquires a plurality of mirror images each including at least a portion of a first mirror, extracted from a plurality of captured images each including a location on a road where a first mirror is installed, and fogging information indicating fogging of at least a portion of the first mirror contained in the plurality of mirror images, and generates a trained model that determines fogging of at least a portion of the second mirror contained in the at least one input mirror image by machine learning using a combination of the acquired plurality of mirror images and the fogging information as training data, when at least one mirror image including at least a portion of a second mirror is input.

Images