Self-position estimation system for estimating the self-position of an unmanned aerial vehicle, flight control system, unmanned aerial vehicle, program, and recording medium

Abstract

Provided is a system for estimating the position of an uncrewed aircraft when the uncrewed aircraft flies within the internal space of a structure, such as a tunnel, having an internal surface on which feature points are difficult to identify.　The self-position estimation system includes: a vertical-section measurement data acquisition unit 138 that acquires vertical-section measurement data measured using a vertical-section measurement device when the uncrewed aircraft flies in the internal space of a structure extending along a prescribed route in the lateral direction; and a vertical-section interior position estimation unit 133 that matches a first prescribed shape corresponding to an inner wall surface of the structure to the vertical-section measurement data acquired by the vertical-section measurement data acquisition unit 138, and estimates the vertical-section interior position of the uncrewed aircraft within the vertical section on the basis of the self position in relation to the matched prescribed shape.

Description

[Technical field]

[0001] The present invention relates to a self-position estimation system that estimates the self-position of an unmanned aerial vehicle, a flight control system, an unmanned aerial vehicle, a program, and a recording medium. [Background technology]

[0002] Recently, attempts have been made to use drones to inspect structures such as tunnels. Drones can estimate their own position using GPS satellite positioning information and fly autonomously along a flight plan route, but it is difficult to obtain accurate position information using GPS in the interior space of structures such as tunnels. Therefore, in such environments, a technology has been proposed that uses Visual SLAM to estimate the self-position as an alternative to GPS-based self-position estimation (see Non-Patent Document 1 below). Here, Visual SLAM is a technology that estimates the self-position and map in parallel by tracking multiple feature points between multiple frames of continuously captured images. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Kenzo Nonami, "All about drone industrial applications", Ohmsha, February 2018, pp.73-79 Summary of the Invention [Problem to be solved by the invention]

[0004] In self-localization using Visual SLAM, feature points are extracted from captured images and the self-localization is estimated based on those feature points. However, there are few feature points on the inner walls of structures such as tunnels, making it difficult to estimate the self-localization with high accuracy using Visual SLAM.

[0005] Therefore, one of the objectives of the present invention is to provide a system and method for estimating the position of an unmanned aerial vehicle when the unmanned aerial vehicle flies within the interior space of a structure, such as a tunnel, whose inner surface makes it difficult to extract feature points. [Means for solving the problem]

[0006] One aspect of the present invention provides a self-location estimation system for estimating the self-location of an unmanned aerial vehicle, wherein the unmanned aerial vehicle includes a longitudinal section measurement device that measures a spatial shape within a longitudinal section in a direction intersecting the flight path of the unmanned aerial vehicle and generates longitudinal section measurement data, and the self-location estimation system includes a longitudinal section measurement data acquisition unit that acquires the longitudinal section measurement data generated by the longitudinal section measurement device when the unmanned aerial vehicle flies through the interior space of a structure extending laterally along a predetermined path, and a longitudinal section position estimation unit that matches the longitudinal section measurement data acquired by the longitudinal section measurement data acquisition unit with a first predetermined shape corresponding to the inner wall surface of the structure, and estimates the position of the unmanned aerial vehicle within the longitudinal section based on the self-location relative to the matched predetermined shape.

[0007] In one aspect of the present invention, the first predetermined shape corresponds at least in part to a geometric shape, and the longitudinal section position estimation unit can match the geometric shape to the longitudinal section measurement data and estimate the position within the longitudinal section based on the self-position relative to the matched geometric shape.

[0008] In one aspect of the present invention, the geometric shape can be a shape composed of convex curves.

[0009] In one aspect of the invention, the geometric shape can be at least one of a partial or complete circle and a partial or complete oval.

[0010] In one aspect of the present invention, the longitudinal section position estimation unit matches a third predetermined shape corresponding to the bottom surface of the structure to the longitudinal section measurement data, and matches the first predetermined shape so that it is located within a predetermined height range with respect to the matched third predetermined shape.

[0011] In one aspect of the present invention, the longitudinal section measurement data is discrete data, and the intra-longitudinal section position estimating unit can remove outliers in the discrete data and perform matching.

[0012] In one aspect of the present invention, the unmanned aerial vehicle may further include an imaging device capable of capturing images of an area below the unmanned aerial vehicle, and may further include an image lateral position estimation unit that estimates the lateral position of the unmanned aerial vehicle from images continuously captured by the imaging device while the unmanned aerial vehicle is flying.

[0013] In one aspect of the present invention, the unmanned aerial vehicle further includes a cross-section measuring device for measuring a spatial shape in a cross-section based on a predetermined direction of the unmanned aerial vehicle and generating cross-section measurement data, and may include a cross-section measurement data acquisition unit for acquiring the cross-section measurement data generated by the cross-section measuring device when the unmanned aerial vehicle flies through an interior space of a structure extending laterally along a predetermined path, and an optical cross-section position / attitude estimation unit for matching the cross-section measurement data acquired by the cross-section measurement data acquisition unit with a second predetermined shape and estimating the position and / or attitude of the unmanned aerial vehicle in the cross-section based on its own position and / or attitude relative to the matched second predetermined shape.

[0014] In one embodiment of the present invention, the second predetermined shape may be at least partially in the form of a line segment or an arc.

[0015] One aspect of the present invention provides a self-location estimation system for estimating the self-location of an unmanned aerial vehicle, wherein the unmanned aerial vehicle includes a cross-section measurement device for measuring a spatial shape in a cross-section based on a predetermined direction of the unmanned aerial vehicle and generating cross-section measurement data, and the self-location estimation system includes a cross-section measurement data acquisition unit that acquires the cross-section measurement data measured by the cross-section measurement device when the unmanned aerial vehicle flies through the interior space of a structure extending laterally along a predetermined path, and an optical cross-section position / attitude estimation unit that matches a second predetermined shape to the cross-section measurement data acquired by the cross-section measurement data acquisition unit, and estimates the cross-section position and / or attitude of the unmanned aerial vehicle based on its self-location and / or attitude relative to the matched second predetermined shape.

[0016] In one aspect of the present invention, the second predetermined shape may be a line segment.

[0017] One aspect of the present invention provides a flight control system for an unmanned aircraft, including the above-mentioned self-location estimation system and a flight control unit that flies the unmanned aircraft along a set flight plan route based on the self-location estimated by the self-location estimation system.

[0018] One aspect of the present invention provides a flight control system for an unmanned aerial vehicle, comprising the above-mentioned self-localization estimation system and a flight control unit that flies the unmanned aerial vehicle along a set flight plan route based on the in-transverse plane position and / or attitude estimated by the self-localization estimation system.

[0019] One aspect of the present invention provides an unmanned aerial vehicle having the above-described self-location estimation system.

[0020] One aspect of the present invention provides an unmanned aerial vehicle having the above-described flight control system.

[0021] One aspect of the present invention provides a method for estimating the self-position of an unmanned aerial vehicle using a computer, the method including: a longitudinal section measurement data acquisition step of acquiring longitudinal section measurement data from a longitudinal section measurement device that measures the spatial shape within a longitudinal section in a direction intersecting the flight path of the unmanned aerial vehicle and generates longitudinal section measurement data when the unmanned aerial vehicle flies through the interior space of a structure extending laterally along a predetermined path; and a longitudinal section position estimation step of matching a first predetermined shape to the longitudinal section measurement data acquired in the longitudinal section measurement data acquisition step, and estimating the position of the unmanned aerial vehicle within the longitudinal section within the longitudinal section based on the self-position relative to the matched predetermined shape.

[0022] One aspect of the present invention provides a method for estimating the self-position of an unmanned aerial vehicle using a computer, the method including: a cross-sectional measurement data acquisition step of acquiring cross-sectional measurement data from a cross-sectional measurement device that measures a spatial shape in a cross-section based on a predetermined direction of the unmanned aerial vehicle and generates cross-sectional measurement data when the unmanned aerial vehicle flies through the interior space of a structure extending laterally along a predetermined path; and a cross-sectional position / attitude estimation step of matching a second predetermined shape to the cross-sectional measurement data acquired in the cross-sectional measurement data acquisition step, and estimating the lateral position and / or attitude of the unmanned aerial vehicle based on its self-position and / or attitude relative to the matched second predetermined shape.

[0023] One aspect of the present invention provides a program for causing a computer to execute the above method.

[0024] One aspect of the present invention provides a computer-readable recording medium having the above-mentioned program recorded thereon. Effect of the Invention

[0025] In accordance with the present invention, the present invention can provide a system and method for estimating the position of an unmanned aerial vehicle when the unmanned aerial vehicle is flying autonomously within the interior space of a structure such as a tunnel. [Brief description of the drawings]

[0026] [Figure 1] FIG. 1 is an external view of a multicopter, which is an example of an unmanned aerial vehicle (multicopter) according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating a flight control system for the unmanned aerial vehicle shown in FIG. [Diagram 3] 2 is a block diagram showing the configuration of an information processing unit of the unmanned aerial vehicle shown in FIG. 1. [Figure 4] 2 is a diagram illustrating a hardware configuration of an information processing unit of the unmanned aerial vehicle shown in FIG. 1. [Figure 5A] 2 is a vertical cross-sectional view in the tunnel width direction showing a tunnel and a flight plan route in one embodiment of the present invention. FIG. [Figure 5B] FIG. 2 is a top view illustrating a tunnel and a flight plan path in one embodiment of the present invention. [Figure 5C] 5B is a cross-sectional view taken along line CC in FIG. 5A, showing a tunnel and a flight plan path in one embodiment of the present invention. [Figure 6] 13 is a diagram for explaining a method for estimating the position of an unmanned aerial vehicle within a longitudinal section. FIG. [Figure 7] 13 is a diagram illustrating a method for estimating the cross-sectional position and heading of an unmanned aerial vehicle within a cross-sectional plane. FIG. [Figure 8] 4 is a flowchart showing the flow of autonomously flying the unmanned aerial vehicle shown in FIG. 1 inside a tunnel and photographing the inner wall surface of the tunnel. [Figure 9] 2 is a flowchart showing in detail the flow of autonomous flight by the unmanned aerial vehicle shown in FIG. 1. [Figure 10] 13 is a diagram showing how straight lines are matched when the tunnel path is curved in cross section. FIG. [Figure 11] FIG. 2 is a diagram showing a longitudinal cross section of the inner wall of a tunnel having protruding and recessed portions. [Figure 12] FIG. 2 shows a longitudinal section of a tunnel with a semicircular upper section and a rectangular lower section. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0027] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. However, the present invention is not limited to the specific embodiment described below, and various embodiments can be taken within the scope of the technical concept of the present invention. For example, the unmanned aerial vehicle of the present invention is not limited to the multicopter shown in FIG. 1, but may be any unmanned aerial vehicle such as a rotorcraft or a fixed-wing aircraft. In addition, the system configuration of the unmanned aerial vehicle 1 is not limited to that shown in the figure, and may be any configuration as long as the same operation is possible. For example, the operation performed by multiple components may be performed by a single component, such as integrating the function of a communication circuit into a flight control unit, or the operation performed by a single component may be performed by multiple components, such as distributing the function of a main calculation unit to multiple calculation units. In addition, various data stored in the memory of the unmanned aerial vehicle 1 may be stored in a different location from the memory, and information recorded in various memories may be stored by distributing one type of information to multiple types, or multiple types of information may be stored together in one type.

[0028] In the following description, an unmanned aerial vehicle autonomously flies through a horizontally extending tunnel and captures images of the inner wall of the tunnel. The shape of the tunnel is known from blueprints, etc., and in this embodiment, the inner space of the tunnel has a semicircular inner shape in a vertical cross section (longitudinal cross section) and extends horizontally (laterally) on the same cross section.

[0029] FIG. 1 is an external view of a multicopter, which is an example of an unmanned aerial vehicle (multicopter) 1 according to an embodiment of the present invention. The unmanned aerial vehicle 1 includes, in terms of appearance, a control unit 101, six motors 102 driven by control signals from the control unit 101, six rotors (rotating wings) 103 that rotate by driving the respective motors 102 to generate lift, six arms 104 connecting the control unit 101 and the respective motors 102, and landing legs 105 that support the unmanned aerial vehicle when landing. The number of motors 102, rotors 103, and arms 104 may each be three or more, such as three or four. The six motors 102 are rotated by control signals from the control unit 101, and the rotation speed of each of the six rotors 103 is thereby controlled, thereby controlling the flight of the unmanned aerial vehicle 1, such as ascent, descent, forward / backward / left / right flight, and turning.

[0030] In addition, a base 106 is attached above the control unit 101. A first camera 108 for photographing an object with high resolution is attached to the base 106 via a support member 107 that rotatably supports the first camera 108.

[0031] In addition, a cross-section measuring device (horizontal section LIDAR) 111 for measuring the shape of the surrounding space in terms of a horizontal section (cross section) is provided in front of the base 106. In this embodiment, the cross-section measuring device 111 is a horizontal section LIDAR, and while the unmanned aerial vehicle 1 is on the ground or hovering, it irradiates a pulsed laser at a predetermined angular interval around the entire circumference around a vertical axis in a horizontal plane, and receives the laser reflected by an object in the surrounding space, thereby measuring the distance to the object in the surrounding space based on the time from irradiation to reception. By driving the cross-section measuring device 111 while the unmanned aerial vehicle 1 is flying in the tunnel, the cross-section measuring device 111 generates cross-section measurement data regarding the horizontal cross-sectional shape (cross-sectional shape) of the inner wall surface of the tunnel (the shape of the tunnel inner space). The cross-section measurement data is, for example, discrete data in polar coordinates based on the front of the unmanned aerial vehicle. Note that, although a horizontal section LIDAR is used as the cross-section measuring device 111 in this embodiment, the present invention is not limited thereto. For example, by capturing images of the entire cross section with a visible light camera centered on the unmanned aerial vehicle, integrating the captured images, and creating a 3D model, it is possible to measure the shape of the inner wall of the tunnel within the cross section and estimate the vehicle's own position within the cross section. SfM is a known algorithm for creating such a 3D model, and Dense Visual SLAM is a known algorithm for estimating the vehicle's own position using SfM.

[0032] A stand 110 is attached below the control unit 101. A second camera (imaging device) 109 is attached below the stand 110 and faces downward in the front for capturing an image of an object for performing Visual SLAM processing, which will be described later.

[0033] In addition, a vertical section measuring device (vertical section LIDAR) 112 for measuring the shape of the surrounding space in a vertical section perpendicular to the central axis is provided in front of the stand 110. In this embodiment, the vertical section measuring device 112 is a vertical section LIDAR, and when the unmanned aerial vehicle 1 is on the ground or hovering, a pulse laser is irradiated at a predetermined angular interval around the entire circumference around a horizontal axis extending in the longitudinal direction in a vertical section of the unmanned aerial vehicle 1, and the laser reflected by an object in the surrounding space is received, thereby measuring the distance to the object in the surrounding space based on the time from irradiation to reception. By driving the vertical section measuring device 112 while the unmanned aerial vehicle 1 is flying in the tunnel, the vertical section measuring device 112 generates vertical section measurement data related to the vertical section shape (vertical section shape) of the inner wall surface of the tunnel. As the vertical section measurement data, for example, discrete data of polar coordinates is obtained. Note that in this embodiment, a vertical section LIDAR is used as the vertical section measuring device 112, but the present invention is not limited to this. For example, the entire circumference of the vertical section centered on the unmanned aerial vehicle is imaged by a visible light camera, the images are integrated, and a 3D model is created, whereby the shape of the inner wall surface of the tunnel in the vertical section can be measured and the self-position in the vertical section can be estimated. SfM is known as an algorithm for creating such a 3D model, and Dense Visual SLAM is known as an algorithm for estimating the self-position using SfM. The unmanned aerial vehicle 1 also has an antenna 117.

[0034] Fig. 2 is a diagram showing a flight control system of the unmanned aerial vehicle shown in Fig. 1. The flight control system 200 of the unmanned aerial vehicle 1 includes a control unit 101, a motor 102 electrically connected to the control unit 101, a rotor 103 mechanically connected to the motor 102, a first camera 108, a second camera 109, a longitudinal section measuring device 112, and a transverse section measuring device 111.

[0035] The control unit 101 is configured to process information for controlling the flight of the unmanned aerial vehicle 1 and to control the electrical signals for that purpose, and is typically a unit configured with circuits required to realize such functions by arranging and wiring various electronic components on a board. The control unit 101 further comprises an information processing unit 120, a communication circuit 121, a control signal generating unit 122, a speed controller 123, and an interface 125.

[0036] The first camera 108 is a camera for capturing an image of an object (in this embodiment, the inner wall of a tunnel) at high resolution. The first camera 108 is attached to the base 106 via a support member 107 that rotatably supports the first camera 108, thereby making it possible to change the imaging direction. The first camera 108 acquires image data of its imaging range while the unmanned aerial vehicle 1 is flying, and the acquired images are stored in a storage device such as a memory described below. The images are typically video images consisting of a series of still images.

[0037] The second camera 109 is a camera for capturing images of an object for performing Visual SLAM processing described later, and is attached to the mount 110 so as to face downward. In order to reduce the load of Visual SLAM processing, the second camera 109, which captures images at a low resolution, is provided separately from the first camera 108. The second camera 109 captures image data within its capture range while the unmanned aerial vehicle 1 is flying. The capture range is set to be continuous. The captured images are stored in the memory of the information processing unit 120. The images are typically moving images consisting of a series of still images.

[0038] Antenna 117 is an aerial for receiving radio signals including information and various data for piloting and controlling unmanned aerial vehicle 1, and for transmitting radio signals including telemetry signals from unmanned aerial vehicle 1.

[0039] The communication circuit 121 is an electronic circuit, typically a wireless signal processing IC, for demodulating the wireless signals received through the antenna 117 to receive control signals, various data, and the like for the unmanned aerial vehicle 1, and inputting the control signals and various data to the information processing unit 120, and for generating wireless signals for carrying telemetry signals and the like output from the unmanned aerial vehicle 1. Note that, for example, the communication of control signals and the communication of control signals and various data may be performed by different communication circuits in different frequency bands. For example, it is possible to adopt a configuration in which communication with a transmitter of a controller (transmitter) for manual operation is performed at a frequency of 950 MHz, and data communication is performed at a frequency of 2 GHz, 1.7 GHz, 1.5 GHz, or 800 MHz.

[0040] Control signal generating unit 122 is configured to convert control command value data obtained by calculation by information processing unit 120 into a pulse signal (such as a PWM signal) representing a voltage, and is typically an IC including an oscillator circuit and a switching circuit. Speed ​​controller 123 is configured to convert the pulse signal from control signal generating unit 122 into a drive voltage that drives motor 102, and is typically a smoothing circuit and an analog amplifier. Although not shown, unmanned aerial vehicle 1 is equipped with a power supply system including a battery device such as a lithium polymer battery or a lithium ion battery, and a power distribution system to each element.

[0041] The interface 125 is a configuration that electrically connects the information processing unit 120 and functional elements such as the first camera 108, the second camera 109, the longitudinal section measuring device 112, and the transverse section measuring device 111 by converting the form of signals so that signals can be transmitted and received between them. For convenience of explanation, the interface is shown as one configuration in the drawings, but it is normal to use a different interface depending on the type of functional element to be connected. Also, depending on the type of signal input / output of the functional element to be connected, the interface 125 may not be necessary. Even in the case of the information processing unit 120 in FIG. 2 that is connected without the interface 125, depending on the type of signal input / output of the functional element to be connected, an interface may be necessary.

[0042] Fig. 3 is a block diagram showing the configuration of the information processing unit of the unmanned aerial vehicle shown in Fig. 1. As shown in Fig. 3, the information processing unit 120 includes a cross-sectional shape data storage unit 130, a flight path data storage unit 131, an image lateral position estimation unit 132, an optical vertical section position estimation unit 133, an optical cross section position / attitude estimation unit 134, a self-position data generation unit 135, a flight control unit 136, a captured image acquisition unit 137, a vertical section measurement data acquisition unit 138, and a cross section measurement data acquisition unit 139. In this embodiment, a self-position estimation system 140 that generates self-position data regarding the self-position of the unmanned aerial vehicle is constituted by a cross-sectional shape data memory unit 130, an image lateral position estimation unit 132, an optical longitudinal section position estimation unit 133, an optical transverse section position / attitude estimation unit 134, a self-position data generation unit 135, a captured image acquisition unit 137, a longitudinal section measurement data acquisition unit 138, and a transverse section measurement data acquisition unit 139.

[0043] Fig. 4 is a diagram showing a hardware configuration of the information processing unit of the unmanned aerial vehicle shown in Fig. 1. Information processing unit 120 includes CPU 120a, RAM 120b, ROM 120c, external memory 120d, input unit 120e, output unit 120f, and communication unit 120g. RAM 120b, ROM 120c, external memory 120d, input unit 120e, output unit 120f, and communication unit 120g are connected to CPU 120a via bus 120h.

[0044] The CPU 120a comprehensively controls each device connected to the system bus 120h. The ROM 120c and the external memory store the BIOS and OS, which are control programs for the CPU 120a, as well as various programs and data required to realize the functions executed by the computer.

[0045] The RAM 120b functions as a main memory, a working area, etc. for the CPU 120a. The CPU 120a loads programs and the like required for executing processes from the ROM 120c or the external memory 120d into the RAM 120b, and executes the loaded programs to realize various operations. The external memory 120d is composed of, for example, a flash memory, a hard disk, a DVD-RAM, a USB memory, or the like.

[0046] The input unit 120e receives operation instructions, etc. from a user, etc. The input unit 120e is configured with input devices, such as, for example, an input button, a keyboard, a pointing device, a wireless remote control, a microphone, and a camera.

[0047] The output unit 120f outputs data processed by the CPU 120a and data stored in the RAM 120b, the ROM 120c, and the external memory 120d. The output unit 120f is composed of output devices such as a CRT display, an LCD, an organic EL panel, a printer, and a speaker.

[0048] The communication unit 120g is an interface for connecting and communicating with an external device directly or via a network, and is configured with an interface such as a serial interface or a LAN interface.

[0049] Each unit 121, 122, 125, 130 to 139 of the flight control system and the position estimation system shown in Figures 2 and 3 is realized by various programs stored in the ROM 120c and the external memory 120d using the CPU 120a, the RAM 120b, the ROM 120c, the external memory 120d, the input unit 120e, the output unit 120f, the communication unit 120g, etc. as resources.

[0050] The cross-sectional shape data storage unit 130 stores cross-sectional shape data relating to the cross-sectional shape of the inner wall surface of the tunnel. The flight route data storage unit 131 stores flight plan route data. The cross-sectional shape data storage unit 130 and the flight route data storage unit 131 are realized by reading data recorded in a recording medium such as the ROM 120c or the external memory 120d and storing the data in the memory. The cross-sectional shape data storage unit 130 and the flight route data storage unit 131 can also be omitted by incorporating the cross-sectional shape data and the flight plan route data into the program.

[0051] In this embodiment, the information processing unit 120 functions as a self-location estimation system and a flight control system, but these systems may be mounted separately from the information processing unit 120 and provided in the unmanned aerial vehicle. The self-location estimation system and the flight control system or components thereof do not need to be configured as one physical device, and may be configured as multiple physical devices. The self-location estimation system and the flight control system may be configured as any appropriate device such as a computer of a ground station separate from the unmanned aerial vehicle, a PC, a smartphone, a tablet terminal, a cloud computing system, or a combination thereof. The functions of each part of the self-location estimation system and the flight control system may be configured to be executed in a distributed manner in one or more of one or more devices provided in the unmanned aerial vehicle and one or more devices separate from the unmanned aerial vehicle.

[0052] Figures 5A to 5C show a tunnel and a flight plan route in one embodiment of the present invention, where Figure 5A is a vertical cross-sectional view in the tunnel width direction, Figure 5B is a top view, and Figure 5C is a CC cross-sectional view in Figure 5A. As shown in Figures 5A to 5C, in this embodiment, the vertical cross-section of the inner surface of the tunnel is semicircular, and extends linearly with the same cross-sectional shape.

[0053] For this reason, the cross-sectional shape data indicating the shape of the inner wall surface 302 of the known tunnel is, for example, 2 +Y 2 =R 2The cross-sectional shape data includes a semicircle (center coordinates (Y=0, Z=0) and radius R) represented by (Z>0) and a line segment (height Z=0 and inclination 0) represented by Z=0 (-R≦Y≦R) corresponding to the tunnel bottom surface 304. Since the shape of the inner wall surface of the tunnel is constant regardless of the depth direction (X direction), X does not exist in the above formula. In this embodiment, the cross-sectional shape data includes data indicating a circle (semicircle) as the first geometric shape corresponding to the inner wall surface of the tunnel, but is not limited to this, and may include data indicating a shape composed of a convex curve such as an ellipse, a rectangle, etc., depending on the cross-sectional shape of the tunnel. Furthermore, the data representing the semicircle includes the radius R and the center coordinates, but may only include information that it is a circle. Furthermore, the data representing the line segment includes the height and inclination, but may only include information that it is a line segment. Note that the geometric shape referred to here refers to a shape that can be specified by a mathematical formula related to the X coordinate, the Y coordinate, and the Z coordinate, for example. The cross-sectional shape data does not necessarily have to include the radius of the semicircle of the inner wall surface of the tunnel, and does not necessarily have to include information about the line segment indicating the bottom surface. Even if it does not include a value related to the radius of the semicircle or data related to the bottom surface, it is possible to perform matching using circles as long as the cross-sectional shape data reflects information that a circular shape is included in part. Furthermore, the cross-sectional shape data does not necessarily have to include the shape in the depth direction (X direction). As will be described later, even if the cross-sectional shape of the tunnel in the depth direction is unknown, autonomous flight along the wall surface is possible.

[0054] The flight plan route data is data that represents a three-dimensional (X, Y, Z) flight plan route of the unmanned aerial vehicle 1, and typically includes a series of multiple waypoints P1 to P2 that exist on the flight plan route. 4n The data is a set of waypoints P1 to P 4nis set in three-dimensional coordinates (X, Y, Z). The flight plan route is typically a collection of straight lines connecting the multiple waypoints in order, but may also be a curve with a predetermined curvature within a predetermined range of the waypoints. The flight plan route data may include data that determines the flight speed at the multiple waypoints. The flight plan route data is typically used to determine the flight plan route in an autonomous flight, but may also be used as a guide during flight in a non-autonomous flight. The flight plan route data is typically input and stored in the unmanned aerial vehicle 1 before flight.

[0055] In this embodiment, the X coordinate is not necessarily limited to a coordinate on a straight line. That is, the X coordinate may be a distance along a path of the tunnel from the starting point toward the depth of the tunnel.

[0056] In this embodiment, as shown in FIGS. 5A to 5C, the flight plan route is set as follows. - At a given position on the vertical cross section (YZ cross section), move a given distance toward the back of the tunnel in the direction of its extension (X direction) (waypoints P1 to P n ), - Move the tunnel center and the concentric circle in the vertical section (YZ section) counterclockwise for a specified angle in Figure 4A (waypoint P n ~P n+1 ), - At a given position on the vertical cross section (YZ cross section), move a given distance toward the hand in the direction of tunnel extension (X direction) (waypoint P n+1 ~P 2n ), - Move the tunnel center and the concentric circle in the vertical section (YZ section) counterclockwise for a specified angle in Figure 4A (waypoint P 2n ~P 2n+1 ), - At a given position on the vertical cross section (YZ cross section), move a given distance toward the back of the tunnel in the direction of its extension (X direction) (waypoint P 2n+1 ~P 3n ), - Move the tunnel center and the concentric circle in the vertical section (YZ section) counterclockwise for a specified angle in Figure 4A (waypoint P 3n ~P 3n+1 ), - At a given position on the vertical cross section (YZ cross section), move a given distance forward in the direction of the tunnel extension (X direction) (waypoint P 3n+1 ~P 4n ) Each waypoint P1~P 4n A position is set for the first camera 108 to capture an image of the inner wall surface of the tunnel.

[0057] The captured image acquisition unit 137 is capable of controlling the first camera 108 and acquires images captured continuously by the first camera 108 . The image lateral position estimation unit 132 estimates the lateral position (X A ,Y A ) is estimated. Visual SLAM (VSLAM), which creates a map and estimates the self-position, is widely known as such an image lateral position estimation unit 132. Various algorithms have been developed for Visual SLAM, such as MonoSLAM and PTAM (Parallel Tracking and Mapping). Image lateral position estimation unit 132 is realized by executing a program that implements Visual SLAM or the like, which is recorded on a recording medium. Note that a stereo camera may be used as second camera 109 instead of a monocular camera, and in this case, self-position and the like can also be estimated by Visual SLAM using a similar principle.

[0058] The longitudinal section measurement data acquisition unit 138 can control the longitudinal section measurement device 112 and acquires the longitudinal section measurement data acquired by the longitudinal section measurement device 112. The optical longitudinal section position estimation unit 133 estimates the longitudinal section position A(Y B , Z B) is estimated. FIG. 6 is a diagram for explaining a method for estimating the position of the unmanned aerial vehicle in the longitudinal section. As shown in FIG. 6, the longitudinal section measurement data consists of points Q1 to Qn (n=16 in the example shown in FIG. 6) expressed in polar coordinates with respect to the position of the longitudinal section measurement device 112 of the unmanned aerial vehicle. In this embodiment, the optical longitudinal section position estimation unit 133 matches a third predetermined shape corresponding to the bottom surface of the tunnel, which in this embodiment is a straight line (line segment), and matches a first predetermined shape corresponding to the inner wall surface of the tunnel, which in this embodiment is a circle (semicircle), so that the center is located within a predetermined height range from this straight line. Note that matching a circle (semicircle) includes not only matching a part of a circular shape, but also naturally matching a circular arc.

[0059] Specifically, first, a straight line is matched to the points Q1 to Qn. As an algorithm for straight line matching, the least square method, RANSAC, etc. can be used, but in this embodiment, an algorithm such as RANSAC that performs matching while excluding outliers is suitable. When matching a straight line using RANSAC, the following procedure is performed. Step 1. First, randomly select two points (or more points) from the discrete points Q1 to Q16. Step 2: Derive a provisional straight line model that passes through the two points determined in step 1. If three or more points are selected in step 1, the model can be determined using the least squares method or the like. Step 3. Calculate the error between the obtained straight line model and the discrete points Q1 to Q16. If the error is smaller than a predetermined value, add it to the "correct model candidates." Step 4. Repeat steps 1 to 3. Step 5. The correct model candidate that best fits the data is taken as the result of line matching.

[0060] In this embodiment, RANSAC is used as an algorithm for performing matching while excluding outliers, but it is also possible to use Markov Localization, Monte-Carlo Localization, Particle Filter, Hough Transform, etc.

[0061] For example, in the example shown in FIG. 6, Q15 is far from the tendency of other discrete points among the discrete points corresponding to the tunnel bottom. In step 1, if Q13 and Q15 are selected, the linear model becomes L1. In contrast, in step 1, if Q12Q16 are selected, the linear model becomes L2. When the error between such linear models L1 and L2 and Q11 to Q16 is calculated, the error of the linear model L1 is larger than the error of the linear model L2. For this reason, in step 5, values ​​that deviate from the tendency of other data, such as Q15 used when setting the linear model L1, are not used as data for determining a line, and a line that removes the influence of outliers, such as the linear model L2, is matched. Similarly, even if points Q1 to Q10 corresponding to other than the bottom of the tunnel are selected in step 1, the inclination of the provisional linear model becomes very large, and it does not become a "correct model candidate" in step 3.

[0062] In this embodiment, the cross-sectional shape data showing the tunnel bottom surface includes data on the inclination and height, and this inclination can also be used to select the "correct model candidate" in step 3. Specifically, for example, if the difference between the inclination or height of the straight line model and the inclination or height of the cross-sectional shape data is equal to or greater than a predetermined value, it should not be determined to be the "correct model candidate."

[0063] Next, a circle is matched to the points Q1 to Qn. The least squares method, RANSAC, etc. can be used as an algorithm for circle matching, but RANSAC, which performs matching by excluding outliers, is suitable, as in line matching. When matching circles using RANSAC, the procedure is as follows. Step 0: Among the discrete points Q1 to Q16, the points below the matched straight line model L2 are excluded from the candidates. Step 1. Randomly select three points from the discrete points Q1 to Q16 remaining in step 0. Step 2. Derive a tentative circular model passing through the three points determined in step 1. Step 3: Calculate the error between the obtained circular model and the discrete points Q1 to Q16. If the error is smaller than a predetermined value, add it to the "correct model candidates." Step 4. Repeat steps 1 to 3. Step 5. The correct model candidate that best matches the data is taken as the result of circle matching.

[0064] For example, in the example shown in FIG. 5, Q3 is far from the trend of other discrete points among the discrete points corresponding to the tunnel bottom. In the example shown in FIG. 5, if Q3, Q4, and Q5 are selected in step 1, the circle model becomes C1. In contrast, if Q1, Q4, and Q8 are selected in step 1, the circle model becomes C2. When the error between such circle models C1 and C2 and Q1 to Q16 is calculated, the error of the circle model C1 is larger than the error of the circle model C2. For this reason, in step 5, values ​​that deviate from the trend of other data such as Q3 used when setting the circle model C1 are not used as data for determining the circle, and a circle from which the influence of the outliers such as the circle model C2 has been removed is matched.

[0065] In this embodiment, in step 0, points below the straight line model L2 are excluded from the candidates, but this is not limiting, and for example, a provisional circular model derived in step 2 whose center does not fall within a predetermined height range may be excluded from the "correct model candidates" in step 3. That is, in this embodiment, if the height Z of the center of the provisional circular model derived in step 2 is outside the predetermined range (-ΔH≦Z≦ΔH), it may be configured to be excluded from the "correct model candidates".

[0066] The optical longitudinal section position estimation unit 133 calculates the position (Y B , Z B In other words, the Y and Z coordinates based on the origin of the circle are calculated as the position in the vertical section (Y B , Z B )

[0067] The cross section measurement data acquisition unit 139 can control the cross section measurement device 111, and acquires the cross section measurement data acquired by the cross section measurement device 111. The optical cross section position / attitude estimation unit 134 estimates the cross section position (distance l from the inner wall surface of the tunnel) and heading θ of the unmanned aerial vehicle in the cross section based on the cross section measurement data acquired by the cross section measurement data acquisition unit 139 and the cross section shape data recorded in the cross section shape data storage unit 130. Note that the heading θ is the angle of the front of the unmanned aerial vehicle based on the X direction in the horizontal plane, and indicates the attitude of the unmanned aerial vehicle.

[0068] Fig. 7 is a diagram for explaining a method for estimating the intra-transverse-plane position and heading of the unmanned aerial vehicle in the transverse plane. As shown in Fig. 7, the transverse-plane measurement data consists of points R1 to Rn (n=16 in the example shown in Fig. 7) expressed in polar coordinates based on the front of the unmanned aerial vehicle, as indicated by the arrow, relative to the position of the transverse-plane measuring device 111 of the unmanned aerial vehicle. In this embodiment, the optical intra-transverse-plane position / attitude estimation unit 134 performs matching using a second predetermined shape, which in this embodiment is a straight line (line segment).

[0069] Specifically, first, a straight line is matched to the points R1 to Rn. As an algorithm for straight line matching, the least square method, RANSAC, etc. can be used, but in this embodiment, RANSAC, which performs matching while excluding outliers, is suitable. When matching a straight line using RANSAC, it is performed as follows. Step 1. First, randomly select two points (or more points) from the discrete points R1 to R16. Step 2: Derive a provisional straight line model that passes through the two points determined in step 1. If three or more points are selected in step 1, the model can be determined using the least squares method or the like. Step 3. Calculate the error between the obtained straight line model and the discrete points R1 to R16. If the error is smaller than a predetermined value, add it to the "correct model candidates." Step 4. Repeat steps 1 to 3. Step 5. The correct model candidate that best fits the data is taken as the result of line matching.

[0070] In this way, the optical cross-sectional position / orientation estimation unit 134 obtains straight lines L3 and L4 corresponding to the tunnel inner wall surface in the cross-sectional plane by matching the discrete points R1 to R16 with straight lines. By determining the distance l to the straight line L3 obtained in this way and the angle θ of the arrow with respect to the straight line L3, the position in the longitudinal section (distance l in the Y direction from the tunnel inner wall surface) and the heading θ are estimated.

[0071] The self-position data generation unit 135 generates self-position data including a three-dimensional position (X coordinates, Y coordinates, Z coordinates) and a posture (heading θ) relative to a reference position. A ,Y A ) and the position in the longitudinal section estimated by the optical position in the longitudinal section estimation unit 133 (Y B , Z B ) and the position in the longitudinal section estimated by the optical transverse section position / orientation estimation unit 134 (distance l in the Y direction from the inner wall surface of the tunnel) and heading θ, to generate self-position data.

[0072] Here, self-location estimation based on an image such as Visual SLAM in a tunnel is prone to errors due to the small number of feature points. Therefore, in this embodiment, the self-location data generator 135 generates self-location data as follows.

[0073] The position in the tunnel depth direction, i.e., the X coordinate, is the lateral position (X A ,Y A ) X coordinate X A will be adopted.

[0074] The position in the tunnel width direction, i.e., the Y coordinate, is the position in the longitudinal section estimated by the optical longitudinal section position estimation unit 133 (Y B , Z B ) Y coordinate B In addition, the Y coordinate may be calculated based on the Y direction distance l from the tunnel inner wall surface estimated by the optical cross-sectional area position / orientation estimation unit 134, or the Y coordinate and the Y B The average of and may be used.

[0075] The height direction position, i.e., the Z coordinate, is the position in the vertical section (Y B , Z B ) Z coordinate Z B will be adopted.

[0076] For the attitude of the unmanned aerial vehicle, i.e., the heading θ, the heading θ estimated by the optical cross-plane position / attitude estimation unit 134 is used.

[0077] In this manner, the self-position data generating unit 135 generates self-position data regarding the self-position (X coordinate, Y coordinate, Z coordinate) and attitude (heading θ) of the unmanned aerial vehicle.

[0078] The flight control unit 136 controls the flight of the unmanned aerial vehicle to follow the flight plan route of the flight plan route data stored in the flight route data storage unit 131 based on the self-position data generated by the self-position data generation unit 135. Specifically, the flight control unit 136 determines the attitude, speed, etc. of the unmanned aerial vehicle 1 using various sensors, determines the current flight position and heading of the unmanned aerial vehicle 1 based on the self-position data generated by the self-position data generation unit 135, calculates a control command value for each rotor 103 by comparing it with target values ​​such as a control signal, a flight plan route (target), a speed limit, and an altitude limit, and outputs data indicating the control command value to the control signal generation unit 122. The control signal generation unit 122 converts the control command value into a pulse signal representing a voltage and transmits it to each speed controller 123. Each speed controller 123 converts the pulse signal into a drive voltage and applies it to each motor 102, thereby controlling the drive of each motor 102 to control the rotation speed of each rotor 103, thereby controlling the flight of the unmanned aerial vehicle 1.

[0079] The flow of how the unmanned aerial vehicle described above autonomously flies inside a tunnel and photographs the inner wall surface of the tunnel will be described below. Fig. 8 is a flow chart showing a flow of autonomously flying inside a tunnel and photographing the inner wall surface of the tunnel by the unmanned aerial vehicle shown in Fig. 1. As shown in Fig. 8, first, the information processing unit 120 refers to the flight plan route data stored in the flight route data storage unit 131, and the flight control unit 136 sets the first waypoint as a target (S100).

[0080] Next, the information processing unit 120 estimates the self-position and attitude of the unmanned aerial vehicle 1 using the self-position data generation unit 135, while the flight control unit 136 controls the number of rotations of the motor 102 based on the estimated self-position and attitude, and the unmanned aerial vehicle 1 performs autonomous flight toward the target (S110). Figure 9 is a flowchart showing in detail the flow of autonomous flight by the unmanned aerial vehicle shown in Figure 1. As shown in FIG. 9, first, cross-section measurement data acquisition unit 139 controls cross-section measuring device 111 to measure cross-section measurement data, and controls cross-section measuring device 111 to acquire the cross-section measurement data (S111: cross-section measurement data acquisition step).

[0081] Next, the optical cross-section position / attitude estimation unit 134 matches lines based on the cross-section shape data recorded in the cross-section shape data memory unit 130 to the cross-section measurement data acquired by the cross-section measurement data acquisition unit 139, and estimates the position (distance l from the tunnel inner wall surface) and attitude (heading θ) of the unmanned aerial vehicle within the cross-section (S112: optical cross-section position / attitude estimation step).

[0082] If the tunnel extends in a straight line and the direction of the heading θ estimated in this manner deviates from the target, the flight control unit 136 calculates control command values ​​for each rotor 103 in order to adjust the heading θ so as to head toward the target, and outputs data indicating the control command values ​​to the control signal generation unit 122 to control flight (S113). Note that in this embodiment, the heading θ is adjusted so as to head toward the target, but this is not limiting, and the heading θ may be adjusted so as to be parallel to the wall surface in the cross section.

[0083] In addition, when the tunnel does not extend in a straight line and the distance along the tunnel path is adopted as the X coordinate, rather than the coordinate along the linear axis, the heading θ is set in a direction along the tunnel wall surface. As a result, even if the path of the tunnel 300 is curved in the cross section as shown in Fig. 10, the straight line L3 (corresponding to L3 in Fig. 7) is matched in a direction approximately equal to the tangent direction of the inner wall surface 302 of the tunnel in the cross section. Therefore, for example, even if the path of the tunnel 300 is unknown, it is possible to fly in the direction of the arrow so as to follow the inner wall surface 302 of the tunnel 300.

[0084] Next, the captured image acquisition unit 137 acquires images continuously captured by the first camera 108 (S114). Then, the image lateral position estimation unit 132 estimates the lateral self-position (X A ,Y A ) is estimated (S115: image lateral position estimation step).

[0085] Next, the optical position in the longitudinal section estimation unit 133 controls the longitudinal section measurement device 112 to measure the longitudinal section measurement data, and controls the longitudinal section measurement device 112 to acquire the longitudinal section measurement data (S116: longitudinal section shape data acquisition step).

[0086] Next, the optical longitudinal section position estimation unit 133 matches circles and lines based on the cross-sectional shape data recorded in the cross-sectional shape data storage unit 130 to the longitudinal section measurement data acquired by the longitudinal section measurement device 112, and calculates the self-position (Y B , Z B ) is estimated (S117: position within longitudinal section estimation step). As described above, since the heading is adjusted in S113, the longitudinal section becomes a section perpendicular to the route, no distortion occurs in the longitudinal section, and matching can be performed more accurately.

[0087] Next, the self-position data generation unit 135 calculates the lateral position (X A ,Y A ) and the position in the longitudinal section estimated by the optical position in the longitudinal section estimation unit 133 (Y B , Z B ), and generates self-position data based on the position in the longitudinal section (distance l in the Y direction from the tunnel inner wall surface) and heading θ estimated by the optical transverse section position / orientation estimation unit 134 (S118: self-position data generation step). Note that the heading θ after adjustment in S113 may be used.

[0088] Then, the flight control unit 136 determines whether the unmanned aerial vehicle 1 has reached the target based on the self-position data generated by the self-position data generation unit 135 (S119).

[0089] If the flight control unit 136 determines that the unmanned aerial vehicle 1 has not reached the target (NO in S119), the flight control unit 136 calculates a control command value for each rotor 103 so as to fly toward the target based on the self-position data generated by the self-position data generation unit 135, and outputs data indicating the control command value to the control signal generation unit 122 (S120). The control signal generation unit 122 converts the control command value into a pulse signal representing a voltage and transmits it to each speed controller 123. Each speed controller 123 converts the pulse signal into a drive voltage and applies it to each motor 102, thereby controlling the drive of each motor 102 to control the rotation speed of each rotor 103, thereby controlling the flight of the unmanned aerial vehicle 1.

[0090] By repeating steps S110 to S120 at predetermined short time intervals, the unmanned aerial vehicle 1 autonomously flies toward the target. Note that, in this embodiment, the estimation of the position and attitude in the transverse plane by S111 and S112, the estimation of the lateral position by S114 and S115, and the estimation of the position in the longitudinal plane by S116 and S117 are performed in this order, but the order of these estimation steps may be changed, or they may be performed simultaneously.

[0091] In this way, when the unmanned aerial vehicle 1 reaches the target, and flight control unit 136 determines in S118 that the unmanned aerial vehicle 1 has reached the target (YES in S119), information processing unit 120 drives first camera 108 to capture an image of the inner wall surface of the tunnel (S130). Image data of the captured image is stored in the memory of information processing unit 120, or recorded in an appropriate terminal via antenna 117.

[0092] And the target reached by unmanned aerial vehicle 1 is the final waypoint P 4nIf not (NO in S140), the next waypoint P is set as the target (S150), and S110 to S140 are repeated.

[0093] In addition, the target reached by unmanned aerial vehicle 1 is the final waypoint P 4n If this is the case (YES in S140), the unmanned aerial vehicle 1 determines that all necessary steps have been completed, and ends the operation by, for example, flying to return to the initial position, via the flight control unit 136.

[0094] According to this embodiment, the following effects are achieved. In the tunnel, GPS and the like cannot be used, and furthermore, a compass cannot be used due to the influence of geomagnetism. In such a case, when estimating the self-position of the unmanned aerial vehicle, it is possible to estimate the position by an algorithm such as Visual SLAM based on an image captured by a camera facing downward. However, even if an attempt is made to estimate the position by an algorithm such as Visual SLAM, there are few feature points on the inner wall surface of a tunnel or the like, making it difficult to estimate the position accurately. In contrast, in this embodiment, the optical longitudinal section position estimation unit 133 matches a circle to the longitudinal section measurement data measured by the longitudinal section measurement device 112, and estimates the position of the unmanned aerial vehicle in the longitudinal section based on the self-position relative to the matched circle. This makes it possible to accurately measure the self-position even in a shape that does not have feature points.

[0095] In this embodiment, the optical longitudinal section position estimating unit 133 matches a geometric shape, particularly a circle formed of a convex curve, to the longitudinal section measurement data, and estimates the longitudinal section position for the matched geometric shape. Since structures such as tunnels require mechanical stability, their cross-sectional shapes are often arc-shaped or arch-shaped. By matching such geometric shapes, particularly circles, it is possible to reduce the amount of calculation and storage, and to estimate an accurate position. Furthermore, matching is possible even if the longitudinal section shape of the tunnel is not specified in advance, as long as there is a geometric shape such as a circle in part.

[0096] 11, for example, in a structure such as a tunnel, there are parts 310 protruding from the wall, such as lighting and ventilation equipment, and parts 320 recessed from the wall, such as evacuation routes. In contrast, in this embodiment, the cross-sectional shape data is discrete data, and the optical longitudinal section position estimation unit 133 removes outliers in the discrete data by the RANSAC algorithm and performs matching, so that it is possible to eliminate the influence of such protruding parts and recessed parts and perform matching.

[0097] In addition, it is difficult to identify the position in the depth direction when flying through the internal space of a structure such as a tunnel where the same cross-sectional shape is continuous. In contrast, in this embodiment, the first camera 108 captures an image of the bottom of the tunnel, and the image lateral position estimation unit 132 estimates the position based on the image captured by the first camera 108, so that the position in the depth direction can be estimated.

[0098] In this embodiment, a straight line (line segment) is matched to the cross-section measurement data measured by the cross-section measuring device 111, and the optical cross-section position / attitude estimation unit 134 estimates the cross-section position (distance from the wall) and heading θ of the unmanned aerial vehicle based on its own position and attitude with respect to the matched straight line. This makes it possible to accurately estimate its own position within the tunnel's internal space. Furthermore, because the cross-section position (distance from the wall) and heading θ of the unmanned aerial vehicle 1 can be estimated, it becomes possible to fly along the wall even if the lateral shape of the tunnel is unknown.

[0099] In this embodiment, the optical cross-sectional position / attitude estimation unit 134 matches a line segment to the cross-sectional measurement data and estimates the position and attitude for the matched line segment. By matching the line segment in this way, even if the tunnel curves laterally, the heading can be adjusted to be approximately parallel to the tangent direction of the tunnel wall, making it possible to fly autonomously along the tunnel wall.

[0100] In this embodiment, the case where the unmanned aerial vehicle flies in a tunnel with a semicircular cross section has been described, but the present invention is not limited to this. For example, as shown in FIG. 12, the present invention can also be applied to a tunnel with a shape in which the upper part 402A of the inner wall surface 402 in the vertical cross section is semicircular and the lower part 402B is rectangular. In such a case, when matching a circle in the vertical cross section, it is preferable to select a "correct model candidate" so that the position based on the bottom surface of the center of the circle (i.e., the straight line matched to the part corresponding to the bottom surface) is located near the position of height b of the lower part 402B upward from the bottom surface (within the range of height b±Δb).

[0101] In this embodiment, the self-position data generating unit 135 calculates the lateral position (X A ,Y A ) and the position in the longitudinal section estimated by the optical position in the longitudinal section estimation unit 133 (Y B , Z B ), and the position in the longitudinal section (distance l in the Y direction from the inner wall surface of the tunnel) and heading θ estimated by the optical transverse section position / attitude estimation unit 134, but the present invention is not limited to this. For example, it is also possible to omit the longitudinal section measuring device 112 and the optical longitudinal section position estimation unit 133. In this case, the lateral position is used for the X coordinate, the Y coordinate and heading θ are calculated from the position in the longitudinal section, and the height (Z coordinate) is determined using an altimeter. With such a configuration, autonomous flight along the tunnel wall is also possible.

[0102] Also, for example, it is possible to omit the cross section measuring device 111 and the optical cross section position / attitude estimation unit 134, in which case the horizontal position is used for the X coordinate, and the vertical section positions are used for the Y coordinate and Z coordinate. The heading θ can also be calculated from the time differential when the unmanned aerial vehicle flies in a specified direction such as forward. With such a configuration, it is possible to accurately estimate the self-position in the tunnel.

[0103] In addition, in this embodiment, the tunnel has a semicircular vertical cross-sectional shape that extends in the longitudinal direction, and a circle is matched to the vertical cross-sectional shape of the tunnel, and a straight line is matched to the cross-sectional shape of the tunnel, but the present invention is not limited to this. That is, if a part of the vertical cross-sectional shape of the tunnel includes a geometric shape such as a circle, an ellipse, or a rectangle, matching based on the geometric shape can be performed. Also, with regard to the cross-sectional shape, if the tunnel extends in an arc shape, and the cross-sectional shape of the tunnel includes a geometric shape, matching based on the geometric shape can be performed. Note that a two-dimensional shape (not including a straight line) is preferable as the shape to be matched in the vertical cross-section.

[0104] In this embodiment, the image lateral position estimation unit 132 estimates the lateral self-position (X A ,Y A ) and based on this, the X coordinate (position in the depth direction) of the unmanned aerial vehicle 1 is determined, but this is not limited to the above, and the X coordinate may also be determined using an IMU.

[0105] In this embodiment, the tunnel extends horizontally, but the present invention can be applied to the tunnel when it is inclined. In such a case, the horizontal direction may be set to be inclined obliquely, or the horizontal direction may be set. Furthermore, the longitudinal section may be set to be a vertical section, or may be a vertical section relative to the horizontal direction that is inclined obliquely. [Explanation of symbols]

[0106] 1 unmanned aerial vehicle 2 Steps 3 Steps 5 Steps 101 Control unit 102 Motor 103 Rotor 104 Arm 105 Landing leg 106 Pedestal 107 Support member 108 First Camera 109 Second Camera 110 Mount 111 Cross-section measuring device 112 Longitudinal section measuring device 117 Antenna 120 Information Processing Unit 121 Communication Circuit 122 Control signal generator 123 Speed ​​Controller 125 Interface 130 Cross-sectional shape data storage unit 131 Flight route data storage unit 132 Image lateral position estimation unit 133 Optical longitudinal section position estimation unit 134 Optical cross-plane position / orientation estimation unit 135 Self-location data generation unit 136 Flight Control Unit 140 Self-location estimation system 200 Flight Control System

Claims

1. A self-location estimation system for estimating the self-location of an unmanned aerial vehicle in a horizontally extending tunnel, comprising: The unmanned aerial vehicle includes a longitudinal section measurement device that measures a spatial shape within a longitudinal section in a direction intersecting a flight path of the unmanned aerial vehicle and generates longitudinal section measurement data; a longitudinal section measurement data acquisition unit that acquires the longitudinal section measurement data generated by the longitudinal section measurement device when the unmanned aerial vehicle flies through an interior space of a structure extending laterally along a predetermined path; a longitudinal section position estimation unit that matches a first predetermined shape corresponding to an inner wall surface of the structure to the longitudinal section measurement data acquired by the longitudinal section measurement data acquisition unit, and estimates a position of the unmanned aerial vehicle in the longitudinal section based on its own position relative to the matched predetermined shape; The unmanned aerial vehicle further includes a cross section measurement device for measuring a spatial shape within a cross section based on a predetermined direction of the unmanned aerial vehicle and generating cross section measurement data; a cross-section measurement data acquisition unit that acquires the cross-section measurement data generated by the cross-section measurement device when the unmanned aerial vehicle flies through an interior space of a structure extending along a predetermined path in the lateral direction; an optical cross-sectional position / attitude estimation unit that matches a second predetermined shape to the cross-sectional measurement data acquired by the cross-sectional measurement data acquisition unit and estimates a cross-sectional position and / or attitude of the unmanned aerial vehicle based on its own position and / or attitude with respect to the matched second predetermined shape; A self-location estimation system comprising:

2. the first predetermined shape corresponds at least in part to a geometric shape; the longitudinal section position estimation unit matches the geometric shape to the longitudinal section measurement data, and estimates the position in the longitudinal section based on a self-position with respect to the matched geometric shape; The self-location estimation system according to claim 1 .

3. The geometric shape is a shape formed by a convex curve. The self-location estimation system according to claim 2 .

4. The geometric shape is at least one of a partial or complete circle and a partial or complete ellipse. The self-location estimation system according to claim 3 .

5. the longitudinal section position estimation unit matches a third predetermined shape corresponding to a bottom surface of the structure to the longitudinal section measurement data; matching the first predetermined shape to be located within a predetermined height range with respect to the matched third predetermined shape; The self-location estimation system according to any one of claims 1 to 4.

6. The longitudinal section measurement data is discrete data; The longitudinal section position estimation unit removes outliers in the discrete data and performs matching. A self-location estimation system according to any one of claims 1 to 5.

7. The unmanned aerial vehicle further includes an imaging device capable of imaging an area below the unmanned aerial vehicle, The image lateral position estimation unit estimates a lateral position of the unmanned aerial vehicle from images continuously captured by the imaging device during flight of the unmanned aerial vehicle. A self-location estimation system according to any one of claims 1 to 6.

8. The second predetermined shape is at least partially in the shape of a line segment or an arc. The self-location estimation system according to claim 1 .

9. The second predetermined shape is a line segment. The self-location estimation system according to claim 1 .

10. A self-location estimation system according to any one of claims 1 to 9, A flight control unit that flies the unmanned aircraft along a set flight plan route based on the self-position estimated by the self-position estimation system. Unmanned aerial vehicle flight control system.

11. An unmanned aerial vehicle comprising the self-location estimation system according to any one of claims 1 to 9.

12. An unmanned aerial vehicle comprising the flight control system of claim 11.

13. A method for estimating the self-position of an unmanned aerial vehicle in a horizontally extending tunnel by a computer, comprising: a longitudinal section measurement data acquisition step of acquiring the longitudinal section measurement data from a longitudinal section measurement device that measures a spatial shape within a longitudinal section in a direction transverse to the flight path of the unmanned aerial vehicle and generates the longitudinal section measurement data when the unmanned aerial vehicle flies through an internal space of a structure extending along a predetermined path in a lateral direction; a longitudinal section position estimation step of matching a first predetermined shape to the longitudinal section measurement data acquired in the longitudinal section measurement data acquisition step, and estimating a position of the unmanned aerial vehicle in the longitudinal section based on its own position with respect to the matched predetermined shape; a cross-sectional measurement data acquisition step of acquiring, from a cross-sectional measurement device that measures a spatial shape within a cross-section based on a predetermined direction of the unmanned aerial vehicle and generates cross-sectional measurement data, the cross-sectional measurement data when the unmanned aerial vehicle flies through an interior space of a structure extending laterally along a predetermined path; a transverse plane position / attitude estimation step of matching a second predetermined shape to the transverse plane measurement data acquired in the transverse plane measurement data acquisition step, and estimating a lateral position and / or attitude of the unmanned aerial vehicle based on its own position and / or attitude relative to the matched second predetermined shape; A self-location estimation method comprising:

14. A program for causing a computer to execute the method according to claim 13.

15. A computer-readable recording medium having the program according to claim 14 recorded thereon.

Images