Positioning device, positioning method, and computer program product

Abstract

The invention provides a positioning device, a positioning method and a computer program product. The position of a moving body is more suitably derived. The positioning device includes: a processing unit that derives a position of a moving body in an actual space on the basis of a light emission position in an image during a first period in which the light emission position of the light emission device in the image obtained by capturing an image of the moving body to which the light emission device is mounted can be identified, and that derives a position of the moving body in the actual space on the basis of the light emission position in the image during a second period different from the first period in which the light emission position in the image cannot be identified; the position of the moving body in the actual space is derived by deriving a light emission position within the image on the basis of a feature pertaining to the appearance of the moving body in the image.

Description

Cross-references to related applications This application claims the priority and benefits of Japanese Patent Application No. 2023-103376 filed on Jun. 23, 2023. The specification, claims, and drawings of Japanese Patent Application No. 2023-103376 are incorporated herein by reference in their entirety. Technical Field The present disclosure relates to a positioning device, a positioning method and a computer program product. Background Art Conventionally, there is known a technology in which a moving object moving in a space derives the position of the device based on a light emission signal of a light emitting device provided in the space (for example, Japanese Patent Application Laid-Open No. 2021-76401). However, the above-mentioned conventional technology has the following problem: if the light emitting device is miniaturized or the distance between the light emitting device and the moving object is increased, it may be impossible to appropriately detect the light emitting signal and derive the position. Summary of the invention An object of the present disclosure is to more appropriately derive the position of a moving object. In order to solve the above-mentioned problems, the positioning device involved in the present disclosure comprises: a processing unit, which, during a first period in which the light-emitting position of the light-emitting device in an image obtained by photographing a moving body equipped with the light-emitting device can be determined, derives the actual spatial position of the moving body based on the light-emitting position in the image; and during a second period different from the first period in which the light-emitting position in the image cannot be determined, estimates the light-emitting position in the image based on features related to the appearance of the moving body in the image, thereby deriving the actual spatial position of the moving body. In order to solve the above-mentioned problems, the positioning device involved in the present disclosure comprises: a processing unit, which, during a first period in which the light-emitting position of the light-emitting device in an image obtained by photographing the light-emitting device by a photographing device installed on the mobile body can be determined, derives the position of the mobile body in actual space based on the light-emitting position of the light-emitting device in the image; and during a second period different from the first period in which the light-emitting position in the image cannot be determined, estimates the light-emitting position of the light-emitting device in the image based on features related to the appearance of the place where the light-emitting device is set in the image, thereby deriving the position of the mobile body in actual space. - Public Effects - According to the present disclosure, the position of a moving object can be derived more appropriately. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 It is a diagram showing the configuration of the positioning system according to the first embodiment. Figure 2 This is a diagram for explaining the principle of positioning in the positioning system according to the first embodiment. Figure 3 It is a block diagram showing the functional structure of the positioning device. Figure 4 This is a diagram showing the contents of light emitting device management data. Figure 5 This is a diagram showing the contents of positioning history data. Figure 6 A diagram showing a photographic image captured by a certain camera at a certain timing. Figure 7 It means that a camera is connected to Figure 6 Figure 2. Photographic images taken at different time intervals. Figure 8 Detailed description of the flowchart of the control procedure of the positioning process. Fig. 9 Detailed description of the process for determining the light emitting position is given in FIG. Fig.10 It is a diagram showing the structure of a positioning system according to a second embodiment. Fig.11 This is a diagram for explaining the principle of positioning in the positioning system according to the second embodiment. Fig.12 It is a block diagram showing the functional structure of a vehicle. Fig.13 It is a diagram showing the contents of light emitting device management data in the second embodiment. Fig.14 It is a diagram showing the contents of positioning history data in the second embodiment. Fig.15 A diagram showing a photographic image captured by a camera at a certain timing. Fig.16 It means that the camera is Fig.15 Figure 2. Photographic images of the same time-lapse image. Fig.17 This is a flowchart showing the control procedure of the positioning process in the second embodiment. Fig.18 This is a flowchart showing the control procedure of the light emitting position determination process in the second embodiment. DETAILED DESCRIPTION Hereinafter, embodiments of the present disclosure will be described based on the drawings. (First embodiment) <Structure of positioning system> The positioning system 1 of this embodiment derives the positions of the subjects M1, M2, ... (hereinafter, referred to as "subject M" when referring to any one subject) in the actual space 2, and records and manages the history of the derived positions of each subject M. In this embodiment, the actual space 2 is a factory site, and the subject M is an operator working in the site. The subject M is equivalent to a mobile object. The positioning system 1 includes: a positioning device 10; a plurality of light emitting devices 30a, 30b, ... (hereinafter, referred to as "light emitting device 30" when referring to any one of the light emitting devices); and a plurality of cameras C1, C2, ... (shooting devices) (hereinafter, referred to as "camera C" when referring to any one of the cameras). Hereinafter, the positions of the subject M and the camera C in the actual space 2 are represented by an XYZ orthogonal coordinate system. The X direction and the Y direction are parallel to the horizontal plane, and the Z direction is the vertical direction upward. In addition, Figure 1 In the embodiment, the positioning device 10 is arranged in the factory, but it is not limited thereto, and the positioning device 10 may also be arranged outside or at a distance from the factory. The light emitting device 30 functions as a position indicator indicating the position of the light emitting device 30 in the actual space 2, and is also called a marker. The light emitting device 30 is installed on the top of the head of the helmet equipped by the subject M, and moves with the subject M. In addition, one light emitting device 30 is installed for one subject M. In addition, the light emitting device 30 can also be installed at a position other than the top of the head as long as the light emission can be visually identified from the outside. The light emitting device 30 has, for example, three LEDs (Light Emitting Diodes) that emit red, green and blue light, and emits light with a unique light emission pattern. That is, each light emitting device 30 changes the light emission color in a time series with a predetermined different pattern for each light emitting device 30. The light emitting device 30 repeatedly performs the action of changing the light emission color according to the light emission pattern. In the following, the length of the period from the start to the end of the light emission pattern is also recorded as a "light emission pattern period". By decoding the light emission pattern of each light emitting device 30, the light emitting device ID (reference) pre-assigned to each light emitting device 30 can be determined. Figure 4 ). In this way, the light emitting device 30 is used in visible light communication for transmitting information (here, the light emitting device ID) through visible light. Each camera C is installed at a given position in the actual space 2. In this embodiment, it is installed on the ceiling of the venue. The camera C can be connected to the positioning device 10 to send and receive data via a wireless or wired communication path. The camera C captures a photographic image 40 (image) at a given frame rate (see Figure 6 as well as Figure 7 ), and the photographic image data 132 (reference Figure 3) is sent to the positioning device 10. Each camera C of this embodiment is installed at a position where it can capture the entire site of the factory. The positioning device 10 analyzes the moving images transmitted from the cameras C to derive the position of each subject person M in the real space 2 . exist Figure 2 1, a light emitting device 30, two cameras C1 and C2, a photographic image 40a taken by the camera C1, and a photographic image 40b taken by the camera C2 at the same time as the photographic image 40a are depicted. Hereinafter, when the photographic images 40a and 40b are not distinguished from each other, they are referred to as "photographic image 40". The position in the photographic image 40 is represented by an xy orthogonal coordinate system set in the photographic image 40. Among them, the x-axis is parallel to the horizontal direction of the matrix arrangement direction of the pixels in the photographic image 40, and the y-axis is parallel to the vertical direction. The positioning device 10 derives the position P3 (X1, Y1, Z1) of the light emitting device 30 (the subject M) in the real space 2 based on the light emitting position Q1 (x1, y1) of the light emitting device 30 in the photographic image 40a, the light emitting position Q2 (x2, y2) of the light emitting device 30 in the photographic image 40b, the position P1 (Xc1, Yc1, Zc1) of the camera C in the real space 2, and the position P2 (Xc2, Yc2, Zc2) of the camera C2 in the real space, by the principle of triangulation. Here, the position P1 of the camera C1 and the position P2 of the camera C2 in the real space 2 are measured in advance and stored in the storage unit 13. When a plurality of light emitting devices 30a, 30b, ... are captured in the photographic images 40a and 40b, each light emitting device 30 is identified based on the light emitting pattern, and the light emitting positions Q1 and Q2 of each light emitting device 30 in the photographic images 40a and 40b are determined. Thereafter, each light emitting device 30 is positioned by the same method as described above. Hereinafter, when an arbitrary light-emitting position is indicated, it will be referred to as “light-emitting position Q.” Furthermore, by using the photographic images 40 of three or more cameras C to derive the position P3 of the subject person M, more accurate positioning can be performed. The positioning device 10 includes a CPU 11 (Central Processing Unit) (processing unit, computer), a RAM 12 (Random Access Memory), a storage unit 13, a communication unit 14, and a bus 15. The various components of the positioning device 10 are connected via the bus 15. In addition, the positioning device 10 may further include an operation unit, a display unit, etc. used by the administrator of the positioning system 1. The CPU 11 is a processor that controls the operation of each part of the positioning device 10 by reading and executing the program 131 stored in the storage unit 13 and performing various calculations. In addition, the positioning device 10 may have multiple processors (e.g., multiple CPUs), and the multiple processors may execute multiple processes executed by the CPU 11 of this embodiment. In this case, the "processing unit" is composed of multiple processors. In this case, the multiple processors may participate in a common process, or the multiple processors may independently and parallelly execute different processes. The RAM 12 provides a memory space for the CPU 11 to work and temporarily stores data. The storage unit 13 is a non-temporary recording medium readable by the CPU 11 of the computer, and stores a program 131 and various data. The storage unit 13 includes, for example, a non-volatile memory such as a HDD (Hard Disk Drive) or a SSD (Solid State Drive). The program 131 is stored in the storage unit 13 in the form of a computer-readable program code. The data stored in the storage unit 13 include, in addition to the photographic image data 132 received from each camera C, light emitting device management data 133, positioning history data 134, and a machine learning model 135. The light emitting device management data 133 stores information related to the plurality of light emitting devices 30 used in the positioning system 1. One data row of the light emitting device management data 133 corresponds to one light emitting device 30. Each data row contains data items such as "light emitting device ID", "light emitting pattern", and "label information". "Light emitting device ID" is a unique symbol assigned to each light emitting device 30. "Light emitting pattern" indicates the light emitting pattern of the light emitting device 30 in the data row. Figure 4 In the example shown, the switching order of the light emission colors of red (R), green (R), and blue (B) is represented by a character string of "R", "G", and "B". "Label information" is metadata added to the teaching data when the teaching data is prepared (annotated) for machine learning in the machine learning model 135 described later. In this embodiment, the subject M who installed the light emitting device 30 of the data row is used as the label information. The positioning history data 134 stores information related to the positioning result of the subject M that is repeatedly performed at each positioning timing. One data row of the positioning history data 134 corresponds to the positioning result of a certain positioning timing. Each data row contains data items of "time", "luminous position in the image", and "positioning result". "Time" indicates the date and time when the positioning of the data row is performed. "Luminous position Q in the image" indicates the luminous position Q (xy coordinates) of the light emitting device 30 in the photographic image 40 of each camera C used in the positioning of the data row. "Positioning result" indicates the position P3 (XYZ coordinates) of the subject M in the actual space 2 derived based on the "luminous position in the image". The machine learning model 135 performs machine learning by including at least a portion of a photographic image 40 of a given subject such as a subject M and teaching data related to the subject. If the teaching data is input, the subject contained in the image of the input teaching data is recognized and the recognition result is output to the CPU 11. In more detail, the machine learning model 135 performs machine learning so that the person contained in the input image is extracted without identifying the individual. Furthermore, by performing additional learning using the photographic image 40 in which the specific subject M is photographed, the specific subject M can be distinguished from other people and extracted (that is, the individual can be recognized). The structure of the machine learning model 135 is not particularly limited, and for example, a neural network or a support vector machine can be used. The machine learning model 135 can be set in an external device different from the positioning device 10. In this case, the CPU 11 of the positioning device 10 causes the machine learning model 135 to perform machine learning or obtain the recognition result of the machine learning model 135 by sending and receiving data between the external device. Figure 3 The communication unit 14 shown performs communication operations in accordance with a predetermined communication standard. For example, the communication unit 14 receives the photographic image data 132 from the camera C through the communication operations. In addition, the communication unit 14 can send positioning results contained in the positioning history data 134 to a predetermined host device not shown. <Positioning system operation> Next, the actions related to positioning by the positioning system 1 in the first embodiment will be described. When the positioning action is started in the positioning system 1, the photographic image 40 is repeatedly photographed by the camera C at a given frame rate, and the photographic image data 132 is sent to the positioning device 10. The frame rate is set so that the interval between photographic images 40 is shorter than the cycle of switching the luminous color in the luminous pattern of the luminous device 30. For example, when the switching cycle of the luminous color in the luminous pattern is about 100 milliseconds, the frame rate of the photographing by the camera C is set to about 20fps. Thus, a dynamic image is photographed by the camera C. Each photographic image 40 is a frame image of the dynamic image. exist Figure 6 The photographic image 40 shown in FIG. 4 shows two subjects M1 and M2. The positioning operation will be described with a focus on subject M1. Figure 6 When the light emitting position Q of the light emitting device 30a installed on the subject M1 is determined in the photographic image 40 of the subject M1, the light emitting position Q is determined. Here, “the light emitting position Q of the light emitting device 30a can be determined” means that the light emission of the light emitting device 30a identified based on the light emission pattern in the dynamic image can be detected in the photographic image 40, and its position in the image can be determined. In detail, in order to identify the light emitting device 30a, from Figure 6 The light pattern of the light emitting device 30 is detected in the dynamic image (multiple photographic images 40) captured within a given period from the time point of capturing the photographic image 40 (just before). The above-mentioned given period is set to be longer than the period of the light pattern of the light emitting device 30. If the light pattern is detected in the dynamic image, the light pattern is compared with Figure 4 Each light pattern in the light-emitting device management data 133 shown. When the pattern detected from the dynamic image matches any light pattern in the light-emitting device management data 133, it is identified as the light-emitting device 30 (here, the light-emitting device 30a) of the light-emitting device ID corresponding to the matching light-emitting pattern. If the light-emitting device 30a is identified, the light-emitting position Q in the photographic image 40 is determined. Figure 6 In FIG. 4 , the radial lines drawn above the light emitting devices 30 a and 30 b indicate that the light emitted by the light emitting devices 30 a and 30 b can be detected in the photographic image 40 and their positions within the image can be determined. Similarly, the light emitting position Q of the light emitting device 30a is also determined in the photographic images 40 of the other cameras C. If the light emitting position Q of the light emitting device 30a is determined in each of the photographic images 40 of the plurality of cameras C (in this embodiment, the cameras C1 and C2), Figure 2 The method shown is used to derive the position P3 of the light emitting device 30a in the real space 2, that is, the position P3 of the subject M1. The position P3 of the subject M2 in the real space 2 is derived by the same operation. Hereinafter, with respect to a certain light-emitting device 30, a period during which the light-emitting position Q of the light-emitting device 30 can be determined from the photographic image 40 will be recorded as the "first period". In addition, with respect to a certain light-emitting device 30, a period during which the light-emitting position Q of the light-emitting device 30 cannot be determined from the photographic image 40 will be recorded as the "second period". The second period is different from the first period. The first period and the second period may be periods corresponding to the time axis of a dynamic image having a plurality of photographic images 40 as frame images, and may not necessarily correspond to real time. That is, in the case of post-analyzing the transition of the position P3 of the subject M based on a dynamic image shot in the past, a part of the playback period of the dynamic image may correspond to the first period, and the other part may correspond to the second period. exist Figure 7 In the photographic image 40 shown, since the subject M1 moves away from the camera C, the light emitting device 30a of the subject M1 cannot be captured by the pixels divided in the middle in the photographic image 40, and the light emitting position Q of the light emitting device 30a cannot be determined. Figure 7 The shooting timing of the photographic image 40 shown corresponds to the second period. Figure 7 In the photographic image 40 shown, the light emitting position Q of the light emitting device 30b can be determined. Figure 7 The imaging timing of the captured image 40 shown corresponds to the first period. In the case where the subject M wears the light emitting device 30 as in the present embodiment, there is a tendency to use a small light emitting device 30 so as not to hinder the movement of the subject M. Figure 7 In this case, when the subject M moves away from the camera C, the light-emitting position Q is likely to be uncertain. In addition, when the surroundings of the light-emitting device 30 are too bright, such as when sunlight directly hits the light-emitting device 30, the light-emitting position Q may be uncertain. In addition, when there is a shield between the light-emitting device 30 and the camera C or when foreign matter is attached to the light-emitting device 30 so that the light of the light-emitting device 30 is temporarily blocked, the light-emitting position Q may also be uncertain. like Figure 7 As shown in FIG. 1 , when the light emitting position Q of the light emitting device 30a in the photographic image 40 cannot be determined, the subject M1 included in the photographic image 40 (photographed in the photographic image 40) is identified based on the features related to the appearance of the subject M1 wearing the light emitting device 30a. More specifically, when the light emitting position Q of the light emitting device 30a cannot be determined, the machine learning model 135 that has been preliminarily machine-learned for identifying the subject M1 is input. Figure 7At least a portion of the photographic image 40 shown is used to identify the subject M1 contained in the photographic image 40 by the machine learning model 135. The machine learning model 135 may input the entire photographic image 40 or extract the vicinity of the light emitting position Q of the light emitting device 30a in the photographic image 40 that is finally determined. In order to recognize the subject M1 of the machine learning model 135, the machine learning model 135 performs machine learning (additional learning) in advance in the first period using teaching data (labeled data) including an image of the subject M1 and label information. As described above, the machine learning model 135 performs machine learning in advance in order to recognize the person included in the input image, but through this additional learning in the first period, the specific subject M1 can be further distinguished from other persons. Specifically, generating Figure 6 The teaching data shown includes a portion of the photographic image 40 corresponding to the first period, which is within a given annotation frame R1 based on the light emitting position Q of the light emitting device 30a, and label information corresponding to the light emitting device 30a, and the teaching data is input to the machine learning model 135 for machine learning. Hereinafter, when referring to an arbitrary annotation frame, it is recorded as "annotation frame R". Tag information in Figure 4 The light emitting device management data 133 shown is previously associated with the light emitting device ID of the light emitting device 30a. Here, the light emitting device ID of the light emitting device 30a is "M001", and the tag information includes information of the subject M1. The tag information may include information related to the appearance characteristics of the subject M1 (height, etc.). When the label information includes the information of the subject M, the annotation frame R is set to a size such that the subject M enters the frame. Figure 6 As shown, the upper side of the rectangular annotation box R is set at a position passing through the light-emitting position Q of the light-emitting device 30, the interval between the left and right sides of the annotation box R is set to be larger than the width of the subject M in the left-right direction, and the interval between the upper and lower sides of the annotation box R is set to be larger than the height of the subject M. For example, the annotation box R is set so that the closer the distance between the position P1 or P2 where the camera C is set and the position P3 in the actual space 2 of the latest subject M derived in the first period, the larger the annotation box R. In the case where the label information includes information on the height of the subject M, the size of the annotation box R can be adjusted based on the information. In addition, the photographic image 40 can also be input into the machine learning model 135 to make it recognize a person, and the annotation box R can be set so as to be a size that includes the recognized person. For example, in the adjustment method in which the annotation box R becomes a similar shape before and after the size adjustment, the size of the annotation box R in the above-mentioned method can be set to the length of the diagonal of the rectangle formed by the annotation box R. In addition, an annotation frame R may be set in the photographic image 40 reduced by a given reduction ratio, and teaching data of the partial image included in the annotation frame R may be input into the machine learning model 135. As described later, when the machine learning model 135 recognizes the subject M, the light emitting device 30 moves away from the camera C, and the subject M captured in the photographic image 40 also becomes smaller. For this reason, by performing machine learning using the reduced photographic image 40 in advance, machine learning can be performed on the basis of appropriately abstracting the fine features that cannot be seen from a distance among the features related to the appearance of the subject M. As a result, the subject M can be easily recognized in the photographic image 40 in which the subject M located far away is reflected. exist Figure 6 In the photographic image 40 shown, in addition to the machine learning of the subject M1, the machine learning of the subject M2 is also performed in parallel. The size of each annotation frame R corresponding to each subject M is independently set according to the size of each subject M in the image. If in Figure 7 In the photographic image 40 shown in the figure, the subject M1 is identified by the machine learning model 135, and the light emitting position Q1 of the light emitting device 30a in the photographic image 40 is estimated based on the position of the identified subject M1 in the photographic image 40 (for example, the center point of the entire subject M1 or the center point of a given part such as the head). In this embodiment, the top of the head of the identified subject M1 is estimated as the light emitting position Q1 of the light emitting device 30a. Once the light emitting position Q1 is estimated, Figure 6 The position P3 of the light emitting device 30a in the actual space 2, that is, the position P3 of the subject M1, is derived using the same method as in the case of FIG. Thus, in the present embodiment, in the photographic images 40 corresponding to the first period with respect to the light emitting device 30 of a certain subject person M among the photographic images 40 captured by the cameras C1 and C2, the light emitting position Q of the light emitting device 30 in the photographic images 40 is derived, and in the photographic images 40 corresponding to the second period with respect to the light emitting device 30 of the subject person M, the light emitting position Q of the light emitting device 30 in the photographic images 40 is estimated based on the recognition result of the machine learning model 135. Then, based on the light emitting position Q derived or estimated for each photographic image 40, the position P3 of the light emitting device 30b in the actual space 2, that is, the position P3 of the subject person M2 is derived. In the above, the subject M included in the photographic image 40 is identified using the machine learning model 135 in the second period, but the present invention is not limited to this, and the subject M may be identified without using the machine learning model 135. That is, in the first period, the features related to the appearance of the subject M may be extracted from the part within the annotation frame R based on the light emitting position Q of the light emitting device 30 in the photographic image 40, and in the second period, the position of the subject M in the photographic image 40 may be determined based on the extracted features. The features related to the appearance of the subject M used here are features that can be distinguished from other subject M, and may be any features as long as they are larger than the light emitting device 30 (features that can be identified even when the light emitting device 30 becomes unrecognizable). For example, when each subject M is equipped with a helmet of a different color, the color of the helmet may be extracted as the feature related to the appearance. In addition, when each subject M wears a different mark (a strap attached to the helmet, an armband, etc.), the mark may be extracted as the feature related to the appearance. In addition, the appearance of the subject M itself, such as the face, may be extracted as a feature. Various known methods can be used as a method of recognizing the subject person M in the photographic image 40 based on the features in the second period. For example, the subject person M may be recognized by performing pattern matching processing using the extracted features in the photographic image 40 . Next, the positioning process executed by the CPU 11 of the positioning device 10 to realize the above-mentioned operation will be described. The positioning process starts when the positioning system 1 is activated. When the positioning process starts, the CPU 11 substitutes the number of cameras C used in the positioning system 1 into a variable N (step S101). The CPU 11 repeatedly determines whether it is a predetermined positioning timing (step S102), and when it is determined to be a positioning timing (step S102 "Yes"), the CPU 11 substitutes 1 into a variable n representing the ordinal number of the camera C (step S103). The CPU 11 obtains the photographic image 40 of the nth camera C (step S104), and executes the light emitting position determination process (step S105). exist Fig. 9 In the following, a process for a certain subject M is described. When there are a plurality of subjects M to be positioned, the light emission position determination process is performed on each subject M in parallel or sequentially. When the light emitting position identification process is called, the CPU 11 determines whether the light emitting pattern of the light emitting device 30 mounted on the subject M is detected from the moving image captured within a given period (immediately before) up to the capturing time of the captured image 40 (step S201 ). When determining that the light emission pattern is detected (Yes in step S201 ), CPU 11 determines that it is the first period. Then, CPU 11 specifies the light emission position Q (xy coordinates) of light emitting device 30 in captured image 40 and records it in positioning history data 134 (step S202 ). Next, the CPU 11 executes steps S203 to S205 for machine learning (additional learning) on ​​the subject M. First, the CPU 11 refers to the light-emitting device management data 133 and obtains the label information corresponding to the light-emitting device ID of the light-emitting pattern determined in step S201 (step S203). Next, the CPU 11 sets the annotation frame R using the above-mentioned method based on the light-emitting position Q of the light-emitting device 30, etc. (step S204). Then, the CPU 11 generates teaching data including a partial image and label information in the annotation frame R and inputs it to the machine learning model 135, so that it performs machine learning for identifying the subject M (step S205). On the other hand, in step S201, when it is determined that the light pattern of the light emitting device 30 installed on the subject M is not detected (step S201 "No"), the CPU 11 determines that it is the second period, and determines whether to perform machine learning (additional learning) on ​​the subject M corresponding to the light emitting device 30 (step S206). When it is determined that machine learning is to be performed (step S206 "Yes"), the CPU 11 inputs the photographic image 40 (or a part thereof) and the label information into the machine learning model 135 to make it recognize and search for the subject M corresponding to the light emitting device 30 (step S207). When the corresponding subject M is identified (step S208 "Yes"), the CPU 11 estimates the light emitting position Q (xy coordinates) in the photographic image 40 of the light emitting device 30 corresponding to the subject M, and records it in the positioning history data 134 (step S209). When step S205 or S209 is completed, or when the decision is “No” in step S206 or S208, CPU 11 ends the light emitting position determination process and returns the process to step S206 or S208. Figure 8 Positioning processing. like Figure 8 The light emitting position determination process (step S105) is completed, and the CPU 11 determines whether the variable n is consistent with the variable N (step S106). If it is determined that the variable n is inconsistent with the variable N (step S106 "No"), the variable n is incremented (step S107), and the process returns to step S104. The light emitting position Q of the light emitting device 30 in the photographic image 40 of each camera is determined or estimated through the loop process of steps S104 to S107. When it is determined that the variable n is consistent with the variable N (step S106: "Yes"), the CPU 11 derives the position P3 of the light emitting device 30 in the actual space 2, that is, the position P3 (XYZ coordinates) of the subject M, based on the positions P1 and P2 of each camera C in the actual space 2 and the light emitting position Q determined or estimated in each photographic image 40 taken by each camera C at the same time (step S108). When there are multiple subjects M to be positioned, the position P3 of each subject M in the actual space 2 is derived. The CPU 11 records the derived position P3 in the positioning history data 134. When step S108 is finished, CPU 11 determines whether to continue positioning (step S109). Here, CPU 11 determines that positioning is to be continued when no instruction to end positioning is given by the manager of positioning system 1. When CPU 11 determines that positioning is to be continued (step S109 "Yes"), it returns the process to step S102. When CPU 11 determines that positioning is not to be continued (step S109 "No"), it ends the positioning process. <Effect> As described above, the positioning device 10 according to the first embodiment includes the CPU 11, which derives the position P3 of the subject M in the actual space 2 based on the light emitting position Q in the photographic image 40 during the first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the subject M equipped with the light emitting device 30 can be determined, and in the second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined, estimates the light emitting position Q in the photographic image 40 based on the features related to the appearance of the subject M in the photographic image 40, thereby deriving the position P3 in the actual space 2 of the subject M. Thus, even when the light emitting position Q of the light emitting device 30 cannot be determined in the photographic image 40 due to an insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, thereby continuing the positioning. Therefore, the position P3 of the subject M can be derived more appropriately. For example, when a small light emitting device 30 is used to be mounted on the subject M, the light emitting position Q may not be determined due to the distance between the camera C and the light emitting device 30. However, according to the structure of the present embodiment, even in such a case, the light emitting position Q can be estimated to continue positioning. By miniaturizing the light emitting device 30, the power consumption can be suppressed, thereby increasing the continuous operation time of the light emitting device 30. In addition, it is easy to mount the light emitting device 30 on the subject M, or it is difficult to cause an unfavorable situation that the light emitting device 30 hinders the movement of the subject M. In addition, since the light emission of the light emitting device 30 is less likely to be blocked than the radio waves used in wireless communication, it is easy to detect, so positioning can be performed more accurately and appropriately than the existing positioning method using radio waves. In addition, by directly determining the light emitting position Q of the light emitting device 30 in the first period, the light emitting position Q is estimated based on the features related to the appearance of the subject M only in the second period, so that high positioning accuracy and more reliable positioning can be achieved. For example, in a method of positioning using only features related to the appearance of the object person M, it is easy to cause erroneous detection of the object person M due to the difference in the appearance of the object person M seen by each camera. In contrast, according to the method of this embodiment using the light-emitting position Q of the light-emitting device 30, such erroneous detection is less likely to occur, and high-precision positioning can be performed. Furthermore, the CPU 11 inputs the teaching data of the part in the annotation frame R based on the light emitting position Q in the photographic image 40 to the machine learning model 135 in the first period, so as to make the machine learning model 135 perform machine learning so as to recognize the subject M contained in the photographic image 40, and in the second period, makes the machine learning model 135 that has performed the machine learning recognize the subject M contained in the photographic image 40, determine the position of the subject M in the photographic image 40 based on the result of the recognition, and estimate the light emitting position Q based on the result of the determination. Thus, since the machine learning can be performed based on the features related to the appearance of the subject M currently being photographed in the first period, the recognition accuracy of the subject M of the machine learning model 135 can be improved. As a result, in the second period, the subject M that has been learned can be distinguished from other subject M by the machine learning model 135, that is, can be identified as an individual. For example, in a factory, a plurality of subject M wearing similar work clothes and helmets can be distinguished from each other and identified. Furthermore, since the comment frame R is determined based on the light emitting position Q of the light emitting device 30 during the learning of the first period, the subject M to be learned can be accurately determined by identifying the light emitting device 30 , thereby generating teaching data. In addition, the CPU 11 may extract features related to the appearance of the subject M in the caption frame R based on the light-emitting position Q in the photographic image 40 during the first period, determine the position of the subject M in the photographic image 40 based on the extracted features during the second period, and estimate the light-emitting position Q based on the result of the determination. Thus, the position of the subject M in the photographic image 40 can be determined during the second period by simpler processing, and the light-emitting position Q of the light-emitting device 30 can be estimated. Furthermore, the CPU 11 sets the annotation frame R so that the closer the distance between the given position P1 or P2 where the camera C that captures the photographic image 40 is set and the position P3 in the actual space 2 of the subject person M derived in the first period, the larger the annotation frame R. Thus, the annotation frame R can be set to an appropriate size corresponding to the number of pixels occupied by the subject person M in the photographic image 40. In addition, the CPU 11 determines the light emitting position Q for each of the plurality of photographic images 40 when the period corresponds to the first period, and estimates the light emitting position Q for each of the plurality of photographic images 40 when the period corresponds to the second period, and derives the position P3 in the actual space 2 of the subject M based on the plurality of light emitting positions Q derived or estimated in the plurality of photographic images 40. Thus, high-precision positioning using the plurality of photographic images 40 can be performed. Furthermore, in the positioning method according to the present embodiment, the CPU 11 estimates the light emitting position Q in the photographic image 40 based on the features related to the appearance of the subject M in the photographic image 40 during the first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the subject M equipped with the light emitting device 30 can be determined, or the position P3 in the actual space 2 of the subject M can be derived based on the light emitting position Q in the photographic image 40. In the second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined, the light emitting position Q in the photographic image 40 is estimated based on the features related to the appearance of the subject M in the photographic image 40, thereby deriving the position P3 in the actual space 2 of the subject M. Thus, even when the light emitting position O of the light emitting device 30 cannot be determined in the photographic image 40 due to an insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, thereby continuing the positioning. Therefore, the position P3 of the subject M can be derived more appropriately. In addition, the program 131 according to the present embodiment causes the CPU 11 to execute the following processing: in a first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the subject M equipped with the light emitting device 30 can be determined, the position P3 of the subject M in the actual space 2 is derived based on the light emitting position Q in the photographic image 40; in a second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined, the light emitting position Q in the photographic image 40 is estimated based on the features related to the appearance of the subject M in the photographic image 40, thereby deriving the position P3 of the subject M in the actual space 2. Thus, even when the light emitting position Q of the light emitting device 30 cannot be determined in the photographic image 40 due to an insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, thereby continuing the positioning. Therefore, the position P3 of the subject M can be derived more appropriately. (Second embodiment) Next, the second embodiment will be described. Hereinafter, differences from the first embodiment will be described, and description of features common to the first embodiment will be omitted. <Structure of positioning system> The positioning system 1 of the present embodiment includes: a vehicle 20 (moving object) having the positioning device 10 and the camera C described above; and a plurality of light emitting devices 30. In the present embodiment, the vehicle 20 moves in an outdoor space as the real space 2. Fig.10The vehicle 20 shown in the figure has two cameras C for photographing the front and rear, but the present invention is not limited thereto. As long as the desired photographing range can be photographed, there may be only one camera C. A plurality of light emitting devices 30 are set at given positions in the actual space 2. The plurality of light emitting devices 30 are arranged in the following configuration: within the range in which the vehicle 20 is likely to move in the actual space 2, no matter where the vehicle 20 is located, the camera C photographs more than two light emitting devices 30. The installation positions of the light emitting devices 30 are measured in advance and recorded in the light emitting device management data 133 (reference Fig.13 ). While the vehicle 20 moves in the actual space 2, the camera C photographs the light emitting device 30. The positioning device 10 mounted on the vehicle 20 derives the position of the vehicle 20 based on the obtained photographic image data 132. In more detail, the positioning device 10 derives the position of the vehicle 20 based on the photographic image 40 captured by the camera C that captures more than two light emitting devices 30, and the position of the photographic image 40 in the actual space 2. In addition, the positioning device 10 may also be provided outside the vehicle 20. In this case, the data of the photographic image 40 of the camera C is sent from the vehicle 20 to the positioning device 10. In addition, in the present embodiment, an outdoor space is exemplified as the actual space 2, but it is not limited to this, and it may also be an indoor space. In addition, the actual space 2 is not limited to the space on the earth, for example, it may also be the space on the moon, etc. exist Fig.11 2 light emitting devices 30a and 30b, 1 camera C, and a photographic image 40 obtained by photographing the light emitting devices 30a and 30b by the camera C are depicted. The positioning device 10 derives the position P4 (X1, Y1, Z1) of the vehicle 20 (camera C) in the actual space 2 based on the light emitting position Q3 (x1, y1) of the light emitting device 30a and the light emitting position Q4 (x2, y2) of the light emitting device 30b in the photographic image 40, and the position P5 (Xa, Ya, Za) of the light emitting device 30a and the position P6 (Xb, Yb, Zb) of the light emitting device 30b in the actual space 2 by the principle of triangulation. Here, the height (Z1) of the camera C is set to be fixed. In addition, the positioning device 10 can also derive the direction (horizontal position) of the vehicle 20 (camera C) based on the position P4 of the vehicle 20 in the actual space 2 and the light emitting positions Q3 and Q4 of the light emitting devices 30a and 30b in the photographic image 40. This method can also derive the position P4 and orientation of the vehicle 20 under the premise that the height of the vehicle 20 (camera C) and the two degrees of freedom (pitch and roll) of the attitude angle are fixed. In addition, when the attitude angle of the vehicle 20 changes due to unevenness of the road surface, the attitude angle detection unit 22 (see Fig.12) is corrected based on the posture angle detection result of the posture angle detection unit 22. For example, the position P4 can be derived based on the correction of the orientation of the photographic image 40 (the light-emitting position of the light-emitting devices 30a and 30b in the photographic image 40) and the like based on the posture angle detection result of the posture angle detection unit 22. Furthermore, by photographing three or more light-emitting devices 30 and using them in the derivation of the position, positioning corresponding to more degrees of freedom can be performed. That is, when two light-emitting devices 30 are used, three-degree-of-freedom positioning of the horizontal position (XY coordinates) and the horizontal orientation can be performed, whereas by using three or more position indicators, four or more degrees of freedom positioning can be performed by adding at least a portion of the two degrees of freedom (pitch and roll) of the vertical direction coordinate (Z coordinate) and the attitude angle. Ideally, by using six light-emitting devices 30, six-degree-of-freedom positioning of the three-dimensional coordinates (horizontal position and altitude) and the three-axis attitude angles (yaw (horizontal orientation), pitch, and roll) can be performed. Furthermore, in Fig.11 In the figure, a case where two light emitting devices 30 are captured by one camera C is illustrated. If a total of more than two light emitting devices 30 are captured in a plurality of photographic images 40 respectively captured by a plurality of cameras C, positioning can be performed by using information on the orientation of each camera C in the same manner as described above. The vehicle 20 includes a positioning device 10, a camera C, a travel drive unit 21, a posture angle detection unit 22, etc. The various components of the vehicle 20 are connected via a communication path 23 such as a bus. The functional configurations of the positioning device 10 and the camera C are the same as those of the first embodiment. However, the contents of the light emitting device management data 133 and the positioning history data 134 stored in the storage unit 13 are partially different from those of the first embodiment. Each data row of the light emitting device management data 133 includes a data item of "actual spatial position of the light emitting device" in addition to "light emitting device ID", "light emitting pattern" and "label information". The contents of "light emitting device ID" and "light emitting pattern" are the same as those in the first embodiment. In this embodiment, information indicating the light emitting device 30 of the data row is recorded as "label information". "Actual spatial position of the light emitting device" is a position (P5, P6, etc.) in the actual space 2 of the light emitting device 30 of the data row measured in advance. The “luminous position Q in the image” in the positioning history data 134 of the present embodiment indicates the luminous position Q (xy coordinates) in the photographic image 40 of each light emitting device 30 captured in the photographic image 40. The “positioning result” indicates the position P4 (XYZ coordinates) of the vehicle 20 in the actual space 2 derived based on the “luminous position in the image”. Fig.12The travel drive unit 21 shown drives (rotates) wheels for running the vehicle 20 , changes the moving direction, and the like in accordance with a control signal sent from the CPU 11 or a control processor (not shown). The posture angle detection unit 22 is, for example, a 6-axis sensor including a 3-axis acceleration sensor for detecting acceleration in the 3-axis direction and a 3-axis angular velocity sensor (gyro sensor) for detecting angular velocity around the 3 axes. The posture angle detection unit 22 derives the posture angle (yaw, pitch, roll) of the vehicle 20 based on the output values ​​of the 3-axis acceleration sensor and the 3-axis angular velocity sensor, and outputs the derives the posture angle (yaw, pitch, roll) of the vehicle 20 to the positioning device 10. <Positioning system operation> Next, the operation related to positioning of the positioning system 1 in the second embodiment will be described. In the present embodiment, when the movement of the vehicle 20 starts, the camera C starts capturing the captured image 40 at the same frame rate as in the first embodiment. exist Fig.15 Two light emitting devices 30a and 30b are captured in the photographic image 40 shown. In the positioning operation, the light emitting position Q in the photographic image 40 is derived or estimated for each of the light emitting devices 30a and 30b by the same method as in the first embodiment. Specifically, when the light emitting device 30 of the light emitting pattern can be identified in the photographic image 40 and the light emitting position Q can be determined, the light emitting position Q is determined. Fig.15 In FIG. 4 , the sawtooth-shaped graphic representation radially drawn around the circle of the light-emitting device 30 indicates that the light-emitting position Q of the light-emitting device 30 can be determined in the photographic image 40. Fig.15 In the example shown, since the light emitting positions Q1 and Q2 of the light emitting devices 30 a and 30 b can both be determined in the photographic image 40 , the light emitting positions Q1 and Q2 are derived from the photographic image 40 . exist Fig.16 In the photographic image 40 shown, the vehicle 20 moves away from the light emitting device 30b, so that the light emitting device 30b cannot be captured with a sufficient number of pixels in the photographic image 40, and the light emitting position Q2 of the light emitting device 30b cannot be determined. Fig.16 The shooting timing of the photographic image 40 shown corresponds to the second period. Fig.16 In the photographic image 40 shown, the light emitting position Q1 of the light emitting device 30a can be identified. Fig.16 The imaging timing of the captured image 40 shown corresponds to the first period. like Fig.16As shown in FIG. 4 , when the light emitting position Q2 of the light emitting device 30b in the photographic image 40 cannot be determined, the place included in the photographic image 40 (photographed in the photographic image 40) is identified based on the features related to the appearance of the place where the light emitting device 30b is installed (installation place, hereinafter referred to as "installation place"). The appearance of the installation place includes information such as the pattern and color of the object where the light emitting device 30 is installed and the object (in this embodiment, the ground) where the light emitting device 30 is installed. In order to easily distinguish the appearance of the installation place of each light emitting device 30, different patterns and colors may be pre-assigned to each light emitting device 30 at the installation place of each light emitting device 30. In this embodiment, when the light emitting position Q2 of the light emitting device 30b cannot be determined, the machine learning model 135 that has been previously machine-learned for identifying the installation location is input. Fig.16 At least a portion of the photographic image 40 shown is used to identify the installation location contained in the photographic image 40 by the machine learning model 135. The machine learning model 135 may input the entire photographic image 40 or extract and input a portion near the light emitting position Q2 of the light emitting device 30b in the photographic image 40 that is finally determined.

[0074] In order to identify the installation location of the machine learning model 135, the machine learning model 135 performs machine learning (additional learning) in advance during the first period using teaching data including an image of the installation location of the light emitting device 30 and label information. Specifically, for example, Fig.15 The teaching data shown, which corresponds to the portion within the given annotation box R2 based on the light-emitting position Q2 of the light-emitting device 30b in the photographic image 40 of the first period, and the label information corresponding to the light-emitting device 30b, are input into the machine learning model 135 for machine learning. When the tag information includes information about the light emitting device 30, the size of the comment frame R is set so that the installation location around the light emitting device 30 is within the frame. Fig.15 As shown, the annotation frame R is set so that the light emitting portion of the light emitting device 30 is located at the center of the rectangular annotation frame R. For example, the annotation frame R is set so that the closer the distance between the position where the light emitting device 30 is set and the latest position P4 of the vehicle 20 in the actual space 2 derived in the first period is, the larger the annotation frame R is. exist Fig.15 In the photographic image 40 shown, in addition to machine learning of the installation location of the light emitting device 30b, machine learning of the installation location of the light emitting device 30a is also performed in parallel. The size of each annotation frame R corresponding to each light emitting device 30 is independently set according to the size of each light emitting device 30 in the image. If in Fig.16In the photographic image 40 shown, the installation location of the light-emitting device 30b is identified by the machine learning model 135, and the light-emitting position Q2 of the light-emitting device 30b in the photographic image 40 is estimated based on the position of the identified installation location in the photographic image 40 (for example, the center point of the entire installation location). Thus, in the present embodiment, the light emitting position Q of the light emitting device 30 in the photographic image 40 is derived for each of the light emitting devices 30 corresponding to the first period among the plurality of light emitting devices 30 captured in the photographic image 40, and the light emitting position Q of the light emitting device 30 corresponding to the second period is estimated based on the recognition result of the machine learning model 135. Then, the position P4 of the vehicle 20 in the actual space 2 is derived based on the derived or estimated light emitting positions Q of the plurality of light emitting devices 30. In the above, in the second period, the machine learning model 135 is used to identify the installation location of the light emitting device 30, but the present invention is not limited to this, and the installation location of the light emitting device 30 may be identified without using the machine learning model 135. That is, in the first period, features related to the appearance of the installation location of the light emitting device 30 may be extracted from the portion within the annotation frame R based on the light emitting position Q of the light emitting device 30 in the photographic image 40, and in the second period, the position of the installation location of the light emitting device 30 in the photographic image 40 may be determined based on the extracted features. The features related to the appearance of the installation location used here may be any features as long as they can be distinguished from other installation locations. Next, the positioning process executed by the CPU 11 of the positioning device 10 in order to realize the above-mentioned operation will be described. When the positioning process starts, the CPU 11 repeatedly determines whether it is the predetermined positioning timing (step S301). If it is determined to be the positioning timing (step S301 "Yes"), the CPU 11 acquires the image 40 captured by the camera C (step S302) and executes the light emitting position determination process (step S303). When the light emitting position specifying process is called, the CPU 11 determines whether the light emitting pattern of the light emitting device 30 is detected from the moving image captured within a given period (immediately before) up to the capturing time of the captured image 40 (step S401 ). When it is determined that at least one light-emitting pattern is detected (step S401: Yes), the CPU 11 determines that it is the first period for each light-emitting device 30 corresponding to the detected light-emitting pattern. In addition, the CPU 11 selects a light-emitting device 30 corresponding to one of the detected light-emitting patterns (step S402). Then, the CPU 11 determines the light-emitting position Q (xy coordinates) of the light-emitting device 30 in the photographic image 40 and records it in the positioning history data 134 (step S403). Next, the CPU 11 executes steps S404 to S406 for machine learning (additional learning) about the installation location of the light emitting device 30. The processing of steps S404 to S406 is similar to that of the CPU 11 except that the learning object M is the installation location. Fig. 9 The processing of steps S203 to S205 in the above is the same, so the description is omitted. If step S406 ends, the CPU 11 determines whether there is an unselected light-emitting device 30 among the light-emitting devices 30 for detecting the light-emitting pattern (step S407). If it is determined that there is an unselected light-emitting device 30 (step S407 "Yes"), the CPU 11 returns the processing to step S402, and determines the light-emitting position Q and performs machine learning for the next light-emitting device 30. When it is determined that there is no unselected light emitting device 30 (step S407 "No"), or when it is determined that the light emitting pattern of the light emitting device 30 is not detected in step S401 (step S401 "No"), the CPU 11 determines whether there is a missing light emitting device 30 (step S408). Here, the missing light emitting device 30 is a light emitting device 30 whose light emitting pattern was detected at any positioning timing in the past and whose light emitting pattern was not detected in the light emitting position determination process being executed. Alternatively, only the light emitting devices 30 whose light emitting patterns were detected within the most recent given period may be used as objects to determine whether there is a missing light emitting device 30 among them. When it is determined that there is a lost light-emitting device 30 (step S408 "Yes"), CPU11 determines that it is the second period with respect to the lost light-emitting device 30. In addition, CPU11 selects one lost light-emitting device 30 (step S409). CPU11 inputs the photographic image 40 (or a part thereof) and the label information into the machine learning model 135 so that it recognizes and searches for the installation location of the lost light-emitting device 30 (step S410). When the corresponding installation location is identified (step S411 "Yes"), CPU11 estimates the light-emitting position Q (xy coordinates) in the photographic image 40 of the light-emitting device 30 corresponding to the identified installation location, and records it in the positioning history data 134 (step S412). When step S412 ends, or when step S411 branches to "No", CPU11 determines whether there is an unselected lost light-emitting device 30 (step S413). If it is determined that there is an unselected lost light emitting device 30 (“Yes” in step S413), the CPU 11 returns the process to step S409 to estimate the light emitting position Q for the next lost light emitting device 30. If it is determined that there is no unselected lost light emitting device 30 (“No” in step S413), or if it is determined that there is no lost light emitting device 30 in step S408 (“No” in step S408), the CPU 11 ends the light emitting position determination process and returns the process to step S409. Figure 8Positioning processing. like Fig.17 After the light emitting position determination process (step S303) is completed, the CPU 11 refers to the light emitting device management data 133 to obtain the position of each light emitting device 30 in the actual space 2 (for example, Fig.11 Then, the CPU 11 derives the position P4 of the camera C in the actual space 2, that is, the position P4 (XYZ coordinates) of the vehicle 20, based on the light emitting position Q of each light emitting device 30 determined or estimated in the photographic image 40 and the positions P5 and P6 of each light emitting device 30 in the actual space 2 (step S305). In addition, the CPU 11 records the derived position P4 in the positioning history data 134. When step S305 is finished, CPU 11 determines whether to continue positioning (step S306). If CPU 11 determines to continue positioning (step S306 "Yes"), it returns the process to step S301, and if CPU 11 determines not to continue positioning (step S306 "No"), it ends the positioning process. <Effect> As described above, the positioning device 10 according to the second embodiment includes the CPU 11, which derives the position P4 of the vehicle 20 in the actual space 2 based on the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the light emitting device 30 by the camera C installed in the vehicle 20 during the first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 can be determined, and derives the position P4 of the vehicle 20 in the actual space 2 by estimating the light emitting position Q of the light emitting device 30 in the photographic image 40 based on the features related to the appearance of the installation location where the light emitting device 30 is installed in the photographic image 40 during the second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined. Thus, even when the light emitting position Q of the light emitting device 30 cannot be determined in the photographic image 40 due to the insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, and positioning can be continued. Therefore, the position P4 of the vehicle 20 can be derived more appropriately. Furthermore, the CPU 11 inputs the teaching data of the portion in the annotation frame R based on the light-emitting position Q in the photographic image 40 to the machine learning model 135 during the first period, thereby causing the machine learning model 135 to perform machine learning so as to recognize the installation location included in the photographic image 40. During the second period, the CPU 11 causes the machine learning model 135 that has performed the machine learning to recognize the installation location included in the photographic image 40, determine the position of the installation location in the photographic image 40 based on the result of the recognition, and estimate the light-emitting position Q based on the result of the determination. Thus, since the machine learning can be performed based on the features related to the appearance of the installation location currently being photographed during the first period, the recognition accuracy of the installation location by the machine learning model 135 can be improved. Furthermore, during the learning during the first period, since the annotation frame R is determined based on the light-emitting position Q of the light-emitting device 30, the installation location of the learning object can be correctly determined by identifying the light-emitting device 30, thereby generating teaching data. In addition, the CPU 11 may extract features related to the appearance of the installation location within the annotation frame R based on the light-emitting position Q in the photographic image 40 during the first period, determine the position of the installation location within the photographic image 40 based on the extracted features during the second period, and estimate the light-emitting position Q based on the result of the determination. Thus, the position of the installation location within the photographic image 40 can be determined during the second period by a simpler process, thereby estimating the light-emitting position Q of the light-emitting device 30. In addition, the CPU 11 sets the annotation frame R so that the closer the distance between the given position where the light emitting device 30 is installed and the position P4 in the actual space 2 of the vehicle 20 derived in the first period is, the larger the annotation frame R is. In this way, the annotation frame R can be set in the photographic image 40 to an appropriate size corresponding to the number of pixels occupied by the installation location. Furthermore, when the photographic image 40 includes a plurality of light-emitting devices 30, the CPU 11 determines the plurality of light-emitting devices 30 based on the light-emitting pattern of each of the plurality of light-emitting devices 30, and when the photographic image 40 includes a plurality of light-emitting devices 30, the CPU 11 discriminates the plurality of light-emitting devices 30 based on the light-emitting pattern of each of the plurality of light-emitting devices 30, discriminates whether each of the plurality of light-emitting devices 30 corresponds to the first period or the second period, determines the light-emitting position Q of the light-emitting device 30 corresponding to the first period, and estimates the light-emitting position Q of the light-emitting device 30 in the photographic image 40 based on the position in the photographic image 40 of the installation location identified based on the features related to the appearance of the installation location for the light-emitting device 30 corresponding to the second period. Thus, high-precision positioning based on the light-emitting positions Q of the plurality of light-emitting devices 30 can be performed. Furthermore, in the positioning method according to the present embodiment, the CPU 11 derives the position P4 of the vehicle 20 in the actual space 2 based on the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the light emitting device 30 by the camera C installed in the vehicle 20 during the first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 can be determined, and in the second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined, the light emitting position Q of the light emitting device 30 in the photographic image 40 is estimated based on the features related to the appearance of the installation location where the light emitting device 30 is installed in the photographic image 40, thereby deriving the position P4 of the vehicle 20 in the actual space 2. Thus, even when the light emitting position Q of the light emitting device 30 cannot be determined in the photographic image 40 due to an insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, thereby continuing the positioning. Therefore, the position P4 of the vehicle 20 can be derived more appropriately. In addition, the program 131 according to the present embodiment causes the CPU 11 to execute the following processing: in a first period in which the light emitting position Q of the light emitting device 30 in the photographic image 40 obtained by photographing the light emitting device 30 by the camera C installed in the vehicle 20 can be determined, the position P4 in the actual space 2 of the vehicle 20 is derived based on the light emitting position Q of the light emitting device 30 in the photographic image 40; in a second period different from the first period in which the light emitting position Q in the photographic image 40 cannot be determined, the light emitting position Q of the light emitting device 30 in the photographic image 40 is estimated based on the features related to the appearance of the installation location where the light emitting device 30 is installed in the photographic image 40, thereby deriving the position P4 in the actual space 2 of the vehicle 20. Thus, even when the light emitting position Q of the light emitting device 30 cannot be determined in the photographic image 40 due to an insufficient number of pixels occupied by the light emitting device 30, the light emitting position Q can be estimated based on the features related to the appearance of the subject M, thereby continuing the positioning. Therefore, the position P4 of the vehicle 20 can be derived more appropriately. (other) In addition, the present disclosure is not limited to the above-mentioned embodiments, and various changes can be made. For example, in the first embodiment, the subject M moving in the factory is exemplified as the moving body of the positioning object, and in the second embodiment, the vehicle 20 moving outdoors is exemplified as the moving body, but it is not limited to these. The positioning system 1 can also use any moving body moving in the actual space 2 as the positioning object. For example, the moving body can also be a forklift moving in a warehouse. In addition, as the light-emitting pattern of the light-emitting device 30, a light-emitting pattern in which the light-emitting colors of three colors, red, green and blue, are changed is illustrated, but it is not limited to this. For example, the color used may be changed, or the number of colors used may be set to less than 2 colors or more than 4 colors. In addition, the light-emitting pattern may be composed of a combination of a light-emitting period of a single light-emitting color and a non-light-emitting period. In addition, the light-emitting colors of a plurality of light-emitting devices 30 may be made different from each other, and the light emission itself under the light-emitting color inherent to each light-emitting device 30 may be used as a light-emitting pattern. In addition, Li-Fi (Light Fidelity) that sends information through high-speed modulated light may also be used. In addition, optical communication using infrared or ultraviolet rays may be used instead of visible light. In addition, in the above description, an example of using a HDD or SSD of the storage unit 13 as a computer-readable medium of the program involved in the present disclosure is disclosed, but the present disclosure is not limited to this example. As other computer-readable media, information recording media such as flash memory and CD-ROM can be used. In addition, as a medium for providing data of the program involved in the present disclosure via a communication line, a carrier wave is also applicable to the present disclosure. Furthermore, it is needless to say that the detailed configuration and detailed operation of each component of the positioning system 1 in the above-described embodiment can be appropriately modified within the scope not departing from the gist of the present disclosure. Although the embodiments of the present disclosure have been described, the scope of the present disclosure is not limited to the above-described embodiments, but includes the scope of the disclosure described in the claims and the scope equivalent thereto.

Claims

1. A positioning device, characterized in that: Equipped with processing department, The processing unit performs the following processing: In a first period in which a light emitting position of the light emitting device in an image obtained by photographing a moving body equipped with the light emitting device can be determined, a position of the moving body in actual space is derived based on the light emitting position in the image, In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position in the image is estimated based on features related to the appearance of the moving object in the image, thereby deriving the position of the moving object in actual space.

2. The positioning device according to claim 1, wherein: The processing unit performs the following processing: During the first period, the machine learning model is caused to perform machine learning so as to recognize the moving object contained in the image by inputting teaching data of a portion within a given frame based on the light-emitting position in the image into the machine learning model. During the second period, the machine learning model that has performed the machine learning recognizes the moving object contained in the image, determines the position of the moving object in the image based on the recognition result, and estimates the light-emitting position based on the determination result.

3. The positioning device according to claim 1, wherein: The processing unit performs the following processing: In the first period, the feature related to the appearance of the moving object within a given frame based on the light emitting position in the image is extracted, In the second period, the position of the moving object in the image is specified based on the extracted feature, and the light emission position is estimated based on the result of the specification.

4. The positioning device according to claim 2 or 3, wherein: The processing unit performs the following processing: The given frame is set so that the given frame becomes larger as the distance between a given position where a photographing device that photographs the image is installed and the position of the moving object in the actual space derived in the first period is closer.

5. The positioning device according to any one of claims 1 to 3, wherein: The processing unit determines the light-emitting position for each of the multiple images when the period corresponds to the first period, and estimates the light-emitting position when the period corresponds to the second period, and derives the position of the moving object in the actual space based on the multiple light-emitting positions derived or estimated from the multiple images.

6. A positioning device, characterized in that: Equipped with processing department, The processing unit performs the following processing: In a first period in which a light emitting position of the light emitting device in an image obtained by photographing the light emitting device by a photographing device mounted on the moving body can be determined, the position of the moving body in actual space is derived based on the light emitting position of the light emitting device in the image, In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position of the light-emitting device in the image is estimated based on features related to the appearance of the place where the light-emitting device is installed in the image, thereby deriving the position of the moving body in actual space.

7. The positioning device according to claim 6, wherein: The processing unit performs the following processing: During the first period, the machine learning model is caused to perform machine learning so as to recognize the place contained in the image by inputting teaching data of a portion within a given frame based on the light-emitting position in the image into the machine learning model. During the second period, the machine learning model that has performed the machine learning recognizes the place contained in the image, the position of the place in the image is determined based on the result of the recognition, and the light-emitting position is estimated based on the result of the determination.

8. The positioning device according to claim 6, wherein: The processing unit performs the following processing: In the first period, the feature related to the appearance of the place within a given frame based on the light-emitting position in the image is extracted, In the second period, the position of the place in the image is specified based on the extracted feature, and the light-emitting position is estimated based on the result of the specification.

9. The positioning device according to claim 7 or 8, wherein: The processing unit performs the following processing: The given frame is set so that the given frame becomes larger as the distance between the given position where the light emitting device is installed and the position of the moving object in the actual space derived in the first period becomes closer.

10. The positioning device according to any one of claims 6 to 8, wherein: The processing unit performs the following processing: When the image includes a plurality of the light emitting devices, the plurality of light emitting devices are identified based on a light emission pattern of each of the plurality of light emitting devices. determining whether each of the plurality of light emitting devices corresponds to the first period or the second period, For the light emitting device corresponding to the first period, determining the light emitting position of the light emitting device, For the light-emitting device corresponding to the second period, the light-emitting position of the light-emitting device in the image is estimated based on the position in the image of the place recognized based on the feature related to the appearance of the place.

11. A positioning method is a positioning method executed by a computer. In a first period in which a light emitting position of the light emitting device in an image obtained by photographing a moving body equipped with the light emitting device can be determined, a position of the moving body in actual space is derived based on the light emitting position in the image, In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position in the image is estimated based on features related to the appearance of the moving object in the image, thereby deriving the position of the moving object in actual space.

12. A positioning method is a positioning method executed by a computer, characterized in that: In a first period in which a light emitting position of the light emitting device in an image obtained by photographing the light emitting device by a photographing device mounted on the moving body can be determined, the position of the moving body in actual space is derived based on the light emitting position of the light emitting device in the image, In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position of the light-emitting device in the image is estimated based on features related to the appearance of the place where the light-emitting device is installed in the image, thereby deriving the position of the moving body in actual space.

13. A computer program product, characterized in that Make the computer execute the following processing: In a first period in which a light emitting position of the light emitting device in an image obtained by photographing a moving body equipped with the light emitting device can be determined, a position of the moving body in actual space is derived based on the light emitting position in the image; In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position in the image is estimated based on features related to the appearance of the moving object in the image, thereby deriving the position of the moving object in actual space.

14. A computer program product, characterized in that Make the computer execute the following processing: In a first period in which a light emitting position of the light emitting device in an image obtained by photographing the light emitting device by a photographing device mounted on the moving body can be determined, deriving a position of the moving body in actual space based on the light emitting position of the light emitting device in the image; In a second period different from the first period in which the light-emitting position in the image cannot be determined, the light-emitting position of the light-emitting device in the image is estimated based on features related to the appearance of the place where the light-emitting device is installed in the image, thereby deriving the position of the moving body in actual space.

Images