Machine learning device and failure judgment device

Abstract

To provide a machine learning device and a failure determination device which can more accurately determine failure of an aircraft warning light in spite of changes in environment surrounding the aircraft warning light.SOLUTION: A machine learning device 10 constructs a learning model for determining failure of an aircraft warning light and comprises: input data acquisition means 111 which acquires, as input data, a photograph obtained by capturing the aircraft warning light; label acquisition means 112 which acquires, as a label, a determination result of a failure state of the aircraft warning light being a subject in the photograph; and learning model construction means 113 which performs supervised learning with a pair of the input data and the label as teacher data to construct a learning model for determining a failure state of the aircraft warning light in a new photograph including the aircraft warning light as a subject.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a machine learning device and a fault determination device. [Background technology]

[0002] The aviation obstruction lights on power transmission towers are set to high brightness during the day and low brightness at night. However, problems with the settings due to lightning strikes and other factors can cause the brightness to be high at night, which can lead to inquiries from residents living near the installation site to the company that installed the aviation obstruction lights. Similarly, aviation obstruction lights on power plant chimneys can also malfunction and stop working.

[0003] For example, Patent Document 1 discloses technology relating to a failure detection system that includes an imaging means for imaging an aviation obstruction light installed at a predetermined position in order to detect a failure of the aviation obstruction light, an imaging control means for controlling the imaging means to adjust at least the direction and focus of the imaging means based on position information relating to the distance and direction from the imaging means to the aviation obstruction light, and obtain a judgment image used to judge whether the aviation obstruction light is faulty, an emission intensity calculation means for calculating a judgment emission intensity of the aviation obstruction light from the judgment image, and a failure judgment means for comparing the judgment emission intensity with a threshold value to judge whether the aviation obstruction light is faulty. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2014-235806 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, in the technology of Patent Document 1, when determining whether an aviation obstruction light is faulty, the determination light intensity obtained from the determination image is compared with a threshold value. However, this does not address the fact that even if the illuminance of the aviation obstruction light itself is the same, the determination light intensity changes depending on the environment surrounding the aviation obstruction light, such as the weather or the time, due to differences in the determination image.

[0006] The present invention aims to provide a machine learning device and a fault determination device that can more accurately determine a fault in an aviation obstruction light even if the environment surrounding the aviation obstruction light changes. [Means for solving the problem]

[0007] In order to achieve the above object, the present invention has the following configuration.

[0008] (1) A machine learning device that constructs a learning model for determining a failure of an aviation obstruction light, the machine learning device comprising: an input data acquisition means that acquires a photograph of the aviation obstruction light as input data; a label acquisition means that acquires as a label a determination result of the failure state of the aviation obstruction light that is the subject of the photograph; and a learning model construction means that constructs a learning model for determining the failure state of the aviation obstruction light in a new photograph that includes the aviation obstruction light as the subject by performing supervised learning using a pair of the input data and the label as teacher data.

[0009] According to (1), it is possible to provide a machine learning device that can more accurately determine whether an aviation obstruction light is faulty even if the environment surrounding the aviation obstruction light changes.

[0010] (2) In the machine learning device described in (1), it is preferable that the learning model construction means constructs one learning model for the multiple aviation obstruction lights.

[0011] According to (2), it is possible to combine learning models that determine the failure of multiple aviation obstruction lights into one.

[0012] (3) In the machine learning device described in (1), it is preferable that the learning model construction means constructs a learning model for each of the multiple aviation obstruction lights.

[0013] According to (3), by constructing a learning model for each aviation obstruction light, it is possible to determine the failure of aviation obstruction lights in more detail.

[0014] (4) A fault judgment device using the learning model constructed by the machine learning device described in any one of (1) to (3), comprising: a fault condition judgment means for determining a fault condition of the aviation obstruction light based on the photograph and the learning model.

[0015] According to (4), it is possible to provide a failure determination device that can more accurately determine a failure of an aviation obstruction light even if the environment surrounding the aviation obstruction light changes. Effect of the Invention

[0016] According to the present invention, even if the environment surrounding the aviation obstruction light changes, it is possible to more accurately determine whether the aviation obstruction light is malfunctioning. [Brief description of the drawings]

[0017] [Figure 1] 1 is a diagram showing an overall configuration of a failure determination system according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a functional block diagram of a learning unit included in the machine learning device according to the embodiment of the present invention. [Diagram 3] 1 is a functional block diagram of a failure determination device according to an embodiment of the present invention; [Figure 4] 4 is a flowchart showing an operation of the machine learning device according to the embodiment of the present invention. [Diagram 5] 4 is a flowchart showing an operation of the failure determination device according to the embodiment of the present invention. [Figure 6]FIG. 13 is a diagram showing an overall configuration of a failure determination system according to a modified example of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] Hereinafter, an embodiment of the present invention will be described with reference to FIGS.

[0019] [Configuration of one embodiment] First, a configuration of a fault determination system 100 according to this embodiment will be described. Fig. 1 is a block diagram showing the fault determination system according to this embodiment. As shown in Fig. 1, the fault determination system 100 includes a machine learning device 10 and a fault determination device 20.

[0020] Here, the machine learning device 10 and the fault determination device 20 are paired one-to-one and connected to be able to communicate with each other. Although not shown in FIG. 1, the machine learning device 10 and the fault determination device 20 may be connected to each other via a network. The network may be, for example, a local area network (LAN), the Internet, a public telephone network, or a combination of these. There are no particular limitations on the specific communication method in the network, whether it is a wired connection or a wireless connection, etc. Alternatively, the machine learning device 10 and the fault determination device 20 may be directly connected via a connection unit, rather than communicating using a network.

[0021] The machine learning device 10 constructs, by supervised learning, a learning model to be used in the fault determination device 20. To this end, the machine learning device 10 includes a learning unit 11 as shown in FIG.

[0022] The learning unit 11 constructs a learning model for determining a failure of an aviation obstruction light. Examples of failures of the aviation obstruction light include excessively high illuminance at night, or conversely, excessively low illuminance (including illuminance of 0 because the aviation obstruction light cannot be turned on in the first place).

[0023] Fig. 2 is a functional block diagram of the learning unit 11. The learning unit 11 includes an input data acquisition unit 111, a label acquisition unit 112, a learning model construction unit 113, and a learning model storage unit 114, as shown in Fig. 2.

[0024] The input data acquisition unit 111 acquires, as input data, a photograph including an aviation obstruction light as a subject. Here, the photograph including an aviation obstruction light as a subject is preferably a photograph taken at a predetermined point, from a predetermined angle, and at a predetermined time (for example, fixed times such as 0:00 AM, 1:00 AM, 2:00 AM, 12:00 PM, 1:00 PM, 2:00 PM, 11:00 PM, etc.). Furthermore, it is preferable that the photographs include photographs taken when the weather at the time the aviation obstruction light was photographed was fine, cloudy, or rainy.

[0025] The photographs as the input data may be a plurality of photographs having one obstacle light as a subject, or may be a plurality of photographs each having a different obstacle light as a subject.

[0026] The label acquisition unit 112 acquires, as a label, the judgment result of the failure state of the aviation obstruction light included as a subject in the above-mentioned photograph. Note that this judgment result is preferably an actual judgment result by the business operator who installed the aviation obstruction light, and is a judgment result expressed as a binary value of whether the aviation obstruction light is normal or abnormal.

[0027] The learning model construction unit 113 performs supervised learning using the pair of the above input data and the above label as teaching data to construct a learning model for determining the fault state of an aviation obstruction light contained in a photograph that includes the aviation obstruction light as a subject, and transmits the constructed learning model to the fault determination device 20.

[0028] In this embodiment, when the above-mentioned input data is a plurality of photographs each having a different aviation obstruction light as a subject, one learning model is constructed for the plurality of aviation obstruction lights.

[0029] The learning model construction unit 113 can be realized, for example, by using a Support Vector Machine (hereinafter also referred to as SVM).

[0030] In this case, the learning model construction unit 113 uses a binarized label indicating whether the obstacle light is normal or abnormal as the label, and calculates a hyperplane that separates the space including the input data so as to maximize the margin as to whether the input data corresponds to the specific fault state. Furthermore, the learning model construction unit 113 can use the coefficients of this hyperplane as parameters of a learning model used by the fault determination device 20 described later for fault state determination.

[0031] The learning model storage unit 114 stores the learning model constructed by the learning model construction unit 113.

[0032] The machine learning device 10 has the above-described configuration and constructs a learning model for determining whether an aviation obstruction light is faulty.

[0033] 3 is a functional block diagram of the failure determination device 20. The failure determination device 20 includes a control unit 21, a storage unit 22, and a display unit .

[0034] The control unit 21 is a part that controls the entire failure determination device 20, and realizes various functions in this embodiment by appropriately reading and executing various programs from a storage area such as a ROM, a RAM, a flash memory, or a hard disk (HDD). The control unit 21 may be a CPU. The control unit 21 includes a photo acquisition unit 211 and a failure state determination unit 212.

[0035] The photo acquisition unit 211 acquires a new photo including an aviation obstruction light as a subject, separately from the photo included in the teacher data used when machine learning was performed by the machine learning device 10. For example, an operator may operate an input device such as a keyboard or a mouse provided in the failure determination device 20 to load a new photo including an aviation obstruction light as a subject into the failure determination device 20, and then the photo acquisition unit 211 may acquire the loaded new photo.

[0036] The failure state determination unit 212 determines the failure state of the aviation obstruction light that is the subject of the photograph acquired by the photograph acquisition unit 211, based on the learning model.

[0037] The storage unit 22 stores the learning model acquired from the machine learning device 10. The storage unit 22 may also store a photo acquired by the photo acquisition unit 211 that includes an aircraft obstruction light as a subject.

[0038] The display unit 23 is a monitor that displays the result of the determination made by the fault state determination unit 212 .

[0039] The failure determination device 20 having the above-described configuration is capable of determining a failure of an aviation obstruction light based on a photograph that includes the aviation obstruction light as a subject.

[0040] [Operation of the second embodiment] Next, a description will be given of the operation of the fault determination system 100 according to this embodiment. First, the operation of the machine learning device 10 will be described with reference to FIG.

[0041] 2.1 Operation of the Machine Learning Device 10 FIG. 4 is a flowchart showing the operation of the learning unit 11. In step S11, the input data acquisition unit 111 acquires, as input data, a photograph including an aircraft obstruction light as a subject.

[0042] In step S12, the label acquisition unit 112 acquires, as a label, the determination result of the failure state of the obstacle light included as a subject in the photograph acquired in step S11.

[0043] In step S13, the learning model construction unit 113 sets the pair of the input data and the label as training data.

[0044] In step S14, the learning model construction unit 113 performs supervised learning using the training data in step S13.

[0045] In step S15, if the machine learning is completed (S15: YES), the process proceeds to step S16. If the machine learning is not completed yet (S15: NO), the process proceeds to step S11.

[0046] In step S16, the learning model construction unit 113 transmits the constructed learning model to the fault determination device 20. After that, all processing ends.

[0047] 2.2 Operation of the failure determination device 20 Next, the operation of the failure determination device 20 will be described with reference to FIG.

[0048] In step S51, the photo acquisition unit 211 acquires a new photo having an aircraft obstruction light as its subject.

[0049] In step S52, the failure state determination unit 212 determines the failure state of the obstacle light included as a subject in the photo acquired by the photo acquisition unit 211, based on the learning model. After that, all processing ends.

[0050] [3. Modifications] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment. Furthermore, the effects described in the present embodiment are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the present embodiment.

[0051] [3.1 Variation 1] In the above embodiment, the machine learning device 10 and the fault determination device 20 are shown as separate entities, but this is not limiting. For example, the machine learning device 10 may be incorporated into the housing of the fault determination device 20, thereby realizing the fault determination system 100 in an integrated manner.

[0052] [3.2 Variation 2] In the above embodiment, one learning model is constructed for a plurality of aviation obstruction lights, but the present invention is not limited to this. Fig. 6 shows a configuration of a fault determination system 100A according to this modified example. As shown in Fig. 6, the machine learning device 10A may be provided with a plurality of learning units (first learning unit 11A to n-th learning unit 11N) each corresponding to each aviation obstruction light, thereby constructing a plurality of learning models each corresponding to each aviation obstruction light. In other words, the machine learning device 10A may construct one learning model for each of a plurality of aviation obstruction lights.

[0053] The fault determination method by the fault determination system 100 is realized by software. When realized by software, the programs constituting this software are installed in a computer. Moreover, these programs may be recorded on removable media and distributed to users, or may be distributed by being downloaded to the user's computer via a network. [Explanation of symbols]

[0054] 10,10A Machine learning device 11 Learning Department 20 Failure determination device 21 Control section 22 Memory section 23 Display section 100,100A fault determination system 111 Input data acquisition unit 112 Label Acquisition Unit 113 Learning Model Construction Department 114 Learning model memory unit 211 Photo Acquisition Department 212 Fault condition determination unit

Claims

1. A machine learning device that constructs a learning model for determining a failure of an aviation obstruction light, An input data acquisition means for acquiring a photograph of the obstacle light as input data; a label acquisition means for acquiring a label as a result of a determination of a failure state of the obstacle light that is a subject of the photograph; a learning model construction means for constructing a learning model for determining a failure state of the aviation obstruction light in a new photograph including the aviation obstruction light as a subject by performing supervised learning using a pair of the input data and the label as training data; Equipped with The input data acquisition means acquires, as input data, photographs of the aircraft obstruction lights when the photographs are taken in clear weather, cloudy weather, and rainy weather, the label acquisition means acquires a determination result of a failure state of the aviation obstruction light for each photograph taken in a sunny, cloudy, or rainy weather at the time when the aviation obstruction light is captured; The learning model construction means is a machine learning device that constructs the learning model by performing supervised learning using as training data a set of photographs taken of the aviation obstruction light when the weather was sunny, cloudy, or rainy, and a judgment result of the failure state of the aviation obstruction light corresponding to the photographs.

2. The machine learning device according to claim 1 , wherein the learning model construction means constructs one learning model for a plurality of the aviation obstruction lights.

3. The machine learning device according to claim 1 , wherein the learning model construction means constructs a learning model for each of the plurality of aviation obstruction lights.

4. A fault determination device using the learning model constructed by the machine learning device according to any one of claims 1 to 3, A fault state determination means for determining a fault state of the aviation obstruction light based on the photograph and the learning model; A failure determination device comprising:

Images