WO2020107248A1 - 一种无人机的安全降落方法、装置、无人机及介质 - Google Patents

一种无人机的安全降落方法、装置、无人机及介质 Download PDF

Info

Publication number
WO2020107248A1
WO2020107248A1 PCT/CN2018/117820 CN2018117820W WO2020107248A1 WO 2020107248 A1 WO2020107248 A1 WO 2020107248A1 CN 2018117820 W CN2018117820 W CN 2018117820W WO 2020107248 A1 WO2020107248 A1 WO 2020107248A1
Authority
WO
WIPO (PCT)
Prior art keywords
return
drone
safe landing
landing point
detected
Prior art date
Application number
PCT/CN2018/117820
Other languages
English (en)
French (fr)
Inventor
李劲松
张立天
Original Assignee
深圳市大疆创新科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2018/117820 priority Critical patent/WO2020107248A1/zh
Priority to CN201880066282.1A priority patent/CN111615677B/zh
Publication of WO2020107248A1 publication Critical patent/WO2020107248A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C11/00—Photogrammetry or videogrammetry, e.g. stereogrammetry; Photographic surveying
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
    • G05D1/10—Simultaneous control of position or course in three dimensions
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B64—AIRCRAFT; AVIATION; COSMONAUTICS
    • B64D—EQUIPMENT FOR FITTING IN OR TO AIRCRAFT; FLIGHT SUITS; PARACHUTES; ARRANGEMENTS OR MOUNTING OF POWER PLANTS OR PROPULSION TRANSMISSIONS IN AIRCRAFT
    • B64D47/00—Equipment not otherwise provided for

Definitions

  • Embodiments of the present invention relate to the field of computer technology, and in particular, to a method, device, drone, and medium for safely landing a drone.
  • agricultural drones may lose GNSS signals due to environmental interference or hardware failures, resulting in the agricultural drones being unable to obtain position information.
  • a binocular vision module under the fuselage so that the agricultural drone can return home based on the position information provided by the binocular vision module.
  • the binocular vision module may not be able to provide a precise position, resulting in a large deviation during the return flight of the drone.
  • the aircraft landed on uneven ground or water on the way back, making the drone lack the guarantee of a safe landing when landing.
  • Embodiments of the present invention provide a method, device, drone, and medium for safely landing an unmanned aerial vehicle, which are helpful to provide a guarantee for the safe landing of an unmanned aerial vehicle.
  • a first aspect of the embodiments of the present invention provides a method for safely landing a drone.
  • the method includes:
  • a second aspect of the embodiments of the present invention provides a safety landing device, which is applied to a drone, and is characterized in that the safety landing device includes a memory and a processor;
  • the memory is used to store program codes
  • the processor calls the program code, and when the program code is executed, it is used to perform the following operations:
  • a third aspect of the embodiments of the present invention provides a drone, including:
  • a power system installed on the fuselage, is used to provide power for the drone;
  • the drone when the drone loses the navigation signal, after determining the return target position and the first return path of the drone, the drone may be controlled to follow the first return path and the return flight
  • the drone When returning to the target position, when it is detected that the current position of the drone is within the first preset range from the returning target position, a safe landing point is detected and recorded, so that the drone can be based on the recorded position Landing at the safe landing point can provide a reliable guarantee for the safe landing of the drone, and at the same time improve the working efficiency of the drone when determining the safe landing point, effectively saving the processing resources of the drone .
  • FIG. 1 is a schematic flowchart of a method for safely landing a drone according to an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of a method for safely landing a drone according to another embodiment of the present invention
  • FIG. 3 is a schematic diagram of an observation area and an area to be detected of a drone provided by an embodiment of the present invention
  • FIG. 4 is a schematic diagram of back-projecting the observation area and the area to be detected as shown in FIG. 3 into a two-dimensional image provided by an embodiment of the present invention
  • FIG. 5 is a schematic diagram of a tilted flying attitude of a drone provided by an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of a safe landing point determined by a drone according to a preset trajectory provided by an embodiment of the present invention
  • FIG. 7 is a schematic flowchart of a method for safely landing a drone according to another embodiment of the present invention.
  • FIG. 8 is a schematic diagram of a return flow process of a drone provided by an embodiment of the present invention.
  • FIG. 9 is a schematic block diagram of a safety landing device provided by an embodiment of the present invention.
  • the navigation signal may be lost due to factors such as environmental interference or hardware failure.
  • the navigation signal of the drone may include at least one of a signal of a positioning sensor or a signal of a compass.
  • the positioning sensor may include a Global Navigation System (GNSS) module installed on the drone.
  • GNSS Global Navigation System
  • agricultural drones that carry out pesticide spraying operations may fail due to environmental factors that interfere with the GNSS module (commonly known as the drones losing stars), or the compass may fail due to hardware failure, resulting in the loss of the drone Navigation signal.
  • the binocular vision module of the drone can be used to perform the visual return.
  • the return target position may be called Home point.
  • the home point can be the starting position of the drone or the position set by the user (it may not be the same as the starting position). When the drone performs a return flight, the home point will be used as the return flight target position and fly toward it.
  • the drone when returning home based on a binocular vision module, the drone may use the starting position of the drone as the return target position of the drone when the navigation signal is lost, so that The drone recorded at the latest before the loss of the navigation signal toward the return target position determines the first return direction of the drone, and determines the drone according to the first return direction and the return target position
  • the first home route further, when the drone detects the missing navigation signal, according to the indication of the first home route, it can return to the home position of the home, when the drone returns to the home
  • the target position perform a safe landing point detection and perform a landing operation when a safe landing point is detected; if a safe landing point is not detected, you need to find a safe landing point and perform a landing.
  • the drone Using the current drone return method, although the safe landing point detection is performed on the current landing position when the return home position is reached, if the detection result indicates that the current landing position is not a safe landing point, the drone needs to re-determine The flight path and the step of re-determining the safe landing point from the new flight path reduces the working efficiency of the drone when determining the safe landing point and wastes the processing resources of the drone.
  • this application proposes a safe landing method for a drone, which can improve the working efficiency of the drone when determining the safe landing point, thereby effectively saving the processing resources of the drone, and at the same time
  • the safe landing of man-machine provides reliable guarantee.
  • FIG. 1 is a schematic flowchart of a method for safely landing a drone according to an embodiment of the present invention. As shown in FIG. 1, the method may include:
  • a positioning failure occurs during the flight of the UAV, it may cause the UAV to lose the navigation signal, for example, the navigation signal is lost when the GNSS of the UAV fails.
  • the UAV detects that the navigation signal is lost it can trigger a binocular vision module under the UAV to perform a visual return, so that the UAV can return to the preset return target position.
  • the preset home return target position is determined from at least one preset home return position, and the preset home return position may be, for example, preset in the drone for performing The location of the battery replacement or flight operation can also be the starting location of the drone.
  • the drone only loses the navigation signal due to the loss of the positioning sensor signal, for example, the positioning sensor fails and the navigation signal is lost, such as the loss of the navigation signal due to the above GNSS failure, you can rely on the compass to determine the drone The first return direction at the time of returning home, so as to determine the first return path of the drone in combination with the determined return home target position and the first returning home direction.
  • the UAV can perform a visual return based on the binocular vision module, and determine its own direction according to the coordinate signal of the positioning sensor during the return.
  • the preset home return target position may be the home return position as described above.
  • the drone can rely on a positioning sensor such as a GNSS module to determine the position of the drone, and determine the first return path of the drone in combination with the determined return target position.
  • the drone can The module recognizes the surrounding environment and locates its position and direction accordingly, and can combine the position and direction information recorded before the loss of the navigation signal to determine the first return path and the return target position of the drone.
  • the positioning device of the drone is normal, and the flight position of the drone will be updated and recorded in real time.
  • the drone can also record the return target position of the drone. Therefore, when the drone is in normal flight, the normally working positioning device can update and record the current position of the drone in real time, and according to the current position and the return target position, update and record the drone toward the return target position in real time direction.
  • the UAV can record the attitude of the UAV in real time through the Inertial Measurement Unit (IMU) on the UAV, so as to record the flight direction of the UAV in real time and obtain the direction of the UAV toward the return target position.
  • IMU Inertial Measurement Unit
  • the UAV loses the navigation signal, for example, when the GNSS module encounters interference loss signal, failure or other problems, it can select any one of the at least one preset return position as the return target position, and can lose the navigation signal
  • the last recorded flight position is determined as the home return position, and the last recorded direction toward the return home position before the loss of the navigation signal is determined as the first home return direction, so that the first home return direction, home target position and home return
  • the starting position determines the first return path of the drone.
  • the first return direction is a direction in which the return home position of the drone points to the return home position.
  • the UAV can determine a straight path as the first return path based on the home position of the home return and the first home return direction, and fly to the home point; it can also be determined based on the home position of the home return, the home point, and the flight position recorded on the way Take the original flight path and use it as the first return path to return to the Home point; of course, the drone can also generate the first return path according to other methods, which is not limited here.
  • S102 Control the drone to perform a return flight based on the first return route and the return destination position.
  • the drone may use any home position recorded in a positioning device such as a GNSS module as the home return target position of the drone.
  • a positioning device such as a GNSS module
  • any home position recorded by the positioning device such as the GNSS module may be It is the position where the drone performs flight operations or battery replacement, or the position where the agricultural drone sprays pesticides, or the position set by the user.
  • the drone can adjust the first return path of the drone in real time in combination with the current position of the drone during the return flight, and control the drone to follow the return flight The indication of the first return route corresponding to each current position is returned to the Home point (ie, the return destination position).
  • the first return path may be basically unchanged, for example, the drone flies directly from the return home position to the return home target location.
  • the first return path may be changed.
  • the UAV encounters obstacles during the return flight and the first return path changes due to the activation of the obstacle avoidance function.
  • the UAV can bypass After the obstacle has finished the obstacle avoidance function, it will return to the starting first return route or continue to return with the changed first return route.
  • the drone When controlling the drone to return home based on the first return path and the return target location, it may also refer to the position information provided by the visual odometer, and the drone may be based on the position information provided by the visual odometer Determine the current location of the drone.
  • the accurate coordinate information of the drone cannot be obtained through GNSS, plus the error of the visual odometer and/or compass during the return flight, may cause no one
  • the aircraft cannot accurately return to the return target position.
  • the final return target position of the drone may be near the return target position of the drone, because the proximity of the return target position may not be suitable for drone safety
  • the landing position so the drone needs to re-determine the safe landing point for the drone to land.
  • the drone can be used during the return flight.
  • the safe landing point is detected at the current position of, and the determined safe landing point is recorded, so that it can land according to the recorded safe landing point in step S104.
  • the drone may start the safety landing point detection when the current position is away from the first preset range of the return target position, and record the detected safety landing point (ie, store the detection Safe landing point), so when the drone is landing at or near the Home point (that is, the home target position), if it is detected that the point is an unsafe landing point, step S104 can be performed according to the recorded safe landing Landing at the point avoids the situation of repeatedly finding a safe landing point, thereby realizing the saving of UAV processing resources and improving the success rate of the drone landing to the safe landing point.
  • the drone can start the safe landing point detection when the current position is within 30 meters of the return target position, and record the detected safe landing point, but it does not land at this time and still flies to the home point .
  • a safe landing point detection may be performed on the current position of the drone, if the drone is detected If the current position is a safe landing point, then directly land, and continue to detect the safe landing point of the position during the landing process. If the current position is detected as an unsafe landing point during the landing process, adjust the flight altitude To perform the height before landing and fly to the recorded safe landing point for landing; if the current position is detected as a safe landing point during landing, continue to reduce the flying altitude until the UAV and the ground are detected When the relative height between them is zero, the power system is turned off, and the drone is landed on the ground.
  • the landing is performed according to the safe landing point recorded within the first preset range from the return home target position.
  • the position information of the safe landing point is recorded. Therefore, when the drone flies toward the recorded safe landing point, the The current position of the man-machine and the recorded position information of the safe landing point determine the second return path of the drone, and the second return path is the path from the current position of the drone to the recorded safety record point.
  • the second return path corresponding to the recorded second return path of the safe landing point of the drone is opposite to the first return path corresponding to the above-mentioned first return path.
  • the direction opposite to the first return direction may be directly used as the second return direction.
  • the drone when the drone loses the navigation signal, after determining the return target position and the first return path of the drone, the drone may be controlled to follow the first return path and the return flight
  • the drone When returning to the target position, when it is detected that the current position of the drone is within the first preset range from the returning target position, a safe landing point is detected and recorded, so that the drone can be based on the recorded position Landing at the safe landing point can provide a reliable guarantee for the safe landing of the drone, and at the same time improve the working efficiency of the drone when determining the safe landing point, effectively saving the processing resources of the drone .
  • S202 Control the UAV to perform a return flight based on the first return route and the return destination position.
  • step S201 and step S202 reference may be made to the description of step S101 and step S102 above, and details are not described herein again.
  • the drone may perform plane detection on the current position of the drone based on binocular vision sensors, and/or perform water surface detection on the current position.
  • the area to be detected for plane detection is determined from the observation area corresponding to the current position in one implementation
  • the observation area of the down-view binocular vision sensor (that is, the binocular vision module) of the UAV is large, as shown in the 301 area (ie, P 1-3D , P 2-3D , P 3- 3D , the area enclosed by P 4-3D ) is assumed to be the observation area of the UAV binocular vision sensor, and when performing plane detection on the current position, it is only necessary to determine the correspondence of the UAV in the physical space It is sufficient if the largest circumscribed circle is flat, and it is not necessary to detect whether all the areas that can be observed by the binocular vision sensor are flat.
  • the area to be detected selected by the drone is too large, it will cause an error in the plane detection result of the current position of the drone, that is, if the largest circumscribed circle corresponding to the current position of the drone should be flat ground, Since the selected area to be detected contains an uneven ground after being enlarged, the drone thinks that the current position is not a plane, and an error occurs in the detection of a safe landing point; and if the area to be detected selected by the drone is too small, It will cause an error in the plane detection result of the current position of the drone, that is, if the current position of the drone is uneven ground, and the selected area to be detected is too small, the selected area to be detected is detected to be flat
  • the ground may cause the UAV to mistake the current position of the unevenness as the flat current position, and control the UAV to descend to the uneven ground to cause a safety failure.
  • the selected area 302 to be detected may be 2 meters*2 meters in physical space To determine whether the current position is a plane according to the plane detection result.
  • the observation area 301 may also be called the detection range of the drone, and the area to be detected 302 may be called Region of interest (ROI) of the image of the UAV.
  • ROI Region of interest
  • the area size of the selected area to be detected 302 is not limited to 2 meters*2 meters, in some cases, it can be determined according to the size of the drone, for example, when the drone is flying on the ground
  • the size of the projection surface is used as the area size of the area to be detected 302, or the size of the projection surface is increased by 50% as the area size.
  • the two-dimensional projection image corresponding to the area to be detected may be determined.
  • the The observation area 301 and the to-be-detected area 302 in the three-dimensional space obtain the observation area image and the to-be-detected area image in the two-dimensional image through the back projection rule.
  • the observation area 301 in the three-dimensional space corresponds to two The observation area image 401 in the two-dimensional image (that is, the area surrounded by P 1-2D , P 2-2D , P 3-2D , and P 4-2D )
  • the area to be detected 302 corresponds to the area to be detected in the two-dimensional image Detection area image 402.
  • the back projection rule may specifically be:
  • P 3d is a three-dimensional space point corresponding to the region to be detected in three-dimensional space
  • P 2d is a two-dimensional image point corresponding to P 3d in two-dimensional space, based on the camera's external parameters (R, T) and the camera's internal parameters K
  • Each three-dimensional space point in the three-dimensional space to be detected can be back-projected according to the back projection rule corresponding to the above formula 2.1, and a two-dimensional projection image 402 corresponding to the to-be-detected region can be determined.
  • the point is the point that needs to process the observation.
  • the points in the two-dimensional projection image need to be converted into three-dimensional space points in the ground coordinate system, and the three-dimensional space points in the ground coordinate system are real Observation point, the height of the raised point represents the height of the object in the actual space above the plane.
  • the resulting point cloud should be concentrated on the same plane; if there is no one
  • the selected area to be detected under the machine includes a tree, then the shape of the point cloud includes a protrusion similar to the top of the tree.
  • the binocular vision module can be used to convert the points in the two-dimensional projection image into three-dimensional space points in the ground coordinate system, that is, the depth map of the binocular vision can be used to convert the area under the aircraft to be detected
  • the physical scene of is converted into a three-dimensional space point set, so that the current position can be plane-detected according to the three-dimensional space point set.
  • RMSCA random sampling consistency algorithm
  • the drone can first obtain the standard plane equation, and calculate the distance between any three-dimensional point in the set of three-dimensional space points and the standard plane equation, and determine the distance based on the distance.
  • the number of interior points in the three-dimensional space point set, the interior points are three-dimensional space points whose distance is less than or equal to a preset distance threshold, and the current point is determined when the number of interior points is greater than or equal to a preset number threshold
  • the location is flat.
  • the plane obtained according to the overdetermined equation is the plane where all the point clouds in the set of points in the three-dimensional space are gathered. If the current position is flat, Then the plane gathered by the point cloud must be fitted to a horizontal plane. If the current position is not flat, the plane gathered by the point cloud must not be fitted to a horizontal plane, which means that the overdetermined equation 2.3 cannot be solved Meaningful parameters.
  • each space point in the three-dimensional space point set can be classified into Point and out point.
  • any space point in the set of three-dimensional space points can be selected as the observation point, and the distance d between the observation point and the plane equation can be calculated, when d is less than or equal to the preset distance
  • the observation point is classified as an inner point; when d is greater than the preset distance threshold, the observation point is classified as an outer point.
  • the coordinates of the observation point are (x, y, z)
  • the distance between the observation point and the plane equation can be calculated according to Equation 2.4:
  • the drone is flying in the flying attitude as shown in FIG. 5 and the safety point is detected during the flight in the flying attitude, the flying attitude and the horizontal plane of the drone There is an angle of inclination between them. Therefore, when the above method is used for plane detection of the current position, the flying height of the drone needs to be measured according to the flying attitude, current position, tilt angle, and flying height H of the drone. Compensate, and perform plane detection on the current position of the drone according to the compensated flying height.
  • the observation area corresponding to the current position may also be obtained from the current position of the drone according to the binocular vision sensor of the drone
  • the area to be detected 302 for water surface detection is determined in 301, and the two-dimensional projection image corresponding to the area to be detected 302 is determined according to the back projection rule of the camera, such as the area 402 in FIG.
  • the machine learning algorithm used for water surface detection is based on the CNN model of the convolutional neural network obtained by grayscale training, that is, the obtained two-dimensional projection image corresponding to the area to be detected can be input into the CNN model, the CNN The model can detect whether there is a water surface in the area corresponding to each frame image in the two-dimensional projection image, and use the detection result as an output to determine whether the current position is a water surface.
  • the current position is a safe landing point
  • the determined safe landing point is recorded.
  • the recorded position can be recorded.
  • the safe landing point corresponds to the relative position of the home target position, wherein the determined safe landing point is a position where the current position is a plane or a non-water surface.
  • other algorithms may also be used to perform plane detection and water surface detection on the current position, for example, using traditional vision-based detection methods to perform plane detection and/or water surface detection.
  • step S205-step S207 when the drone is within the first preset range from the return target position, a safe landing point detection can be performed on the current position and the available drone landing can be recorded Safe landing point, when the UAV continues to fly to the second preset range from the return target position, the second preset range from the current position to the return target position is the return target position Nearby, where the first preset range is greater than the second preset range, the first preset range may be, for example, 20 meters or 10 meters, and the second preset range may be, for example, 5 meters or 10 meters and so on.
  • step S206 is performed according to the recorded Steps for landing at a safe landing point.
  • the second return path can be determined first according to the current position of the drone and the recorded safe landing point, and the first position can be determined based on the current position and the recorded safe landing point.
  • the second home route may be directed from the current position (that is, a position near the home target position) to the direction of the recorded safe landing point, and may further be based on the second home route Path, flying toward the recorded safe landing point, and when the binocular vision sensor determines that the current position of the drone is the recorded safe landing point, controls the drone to land.
  • step S207 is performed to perform the landing step according to the detected safe landing point.
  • the drone is in the landing process . It is necessary to continuously detect whether the detected safe landing point is safe, because the drone is in the second preset range from the return home position, usually the user replaces the battery or the roadside, so no one When the aircraft detects that the current position is a safe landing point and makes a landing, it is likely that someone will appear. If the safe landing point is not continuously detected during the descent process, the drone may also land at an unsafe position. In case of safety failure.
  • the UAV When the UAV continuously detects that the detected safe landing point is safe, it successfully lands to the detected safe landing point; or, if it is detected that the safe landing point is unsafe during the landing process, it will The current flying height of the drone is adjusted to a preset flying height, and the preset flying height is the flying height when the drone returns to home (that is, the height before landing), and executes Steps for landing at a safe landing point.
  • the preset flying height is set to ensure that the binocular vision module under the drone is in a higher accuracy range during the return flight, and the preset flying height may be 2 meters or 2.5 meters, for example.
  • a safe landing point detection may be performed according to a preset trajectory within a third preset range from the home target position, where the preset trajectory includes a spiral trajectory, a polyline trajectory, or a linear trajectory
  • the broken-line trajectory may be, for example, a “Z”-shaped trajectory, so that when a safe landing point is detected in the preset trajectory, the landing can be based on the detected safe landing point.
  • the drone returns from point B (ie, the home position of the return flight) to point A (ie, the home position of the return flight) as shown in FIG. 6, it is near the target position (the current position of the drone)
  • point D the current position of the drone
  • the UAV does not record a safe landing point in the first preset range, it can follow the preset trajectory, for example, it can be a gray spiral trajectory as shown in the figure to find a safe landing point from point D, if the spiral If a safe landing point is detected at the K point of the shaped trajectory, it will land at the K point.
  • the UAV when the UAV does not detect a safe landing point within a second preset range from the return target position during the return flight, and determines the first pre-return distance from the return target position There is no safe landing point recorded in the set range, and the drone can also be controlled to hover at the return target position to wait for the user's operation instructions. The drone performs the landing.
  • a safe landing point detection may be performed on the return target position If it is detected that the home return target position is a safe landing point, then landing is performed at the home return target position, and if it is detected that the home return target location is not a safe landing point, then landing in step S206 is performed according to the recorded safe landing point A step of.
  • the drone if the drone loses the navigation signal, after determining the return target position and the first return path of the drone, the drone can be controlled to return based on the first return path and the return target position, and
  • performing plane detection and water surface detection on the current position according to a preset plane detection algorithm and a preset water surface detection algorithm is as follows:
  • the return and landing process of the drone provides safety guarantee, which can effectively reduce the possibility of a safety failure of the drone after landing, because the safe landing point determined as a plane and/or the safe landing point on the non-water surface are recorded,
  • the UAV returns to the second preset range from the return target position, it can perform a safe landing point detection and make a safe landing attempt.
  • the landing will be based on the recorded safe landing point. If a safe landing point is detected, the landing will be carried out directly, which can increase the speed at which the UAV determines the safe landing point, thereby increasing Man-machine safe landing speed.
  • an application scenario diagram of a safe landing method based on a drone is proposed.
  • the drone takes off from point A and flies to B according to the navigation signal provided by GNSS.
  • the gray curve in Figure 7 identifies the flight path of the drone from point A to point B. During the flight of the drone from point A to point B, it will be refreshed in real time according to the reliable coordinate information provided by GNSS And record the location information of the drone.
  • the UAV has a GNSS failure at point B and the navigation signal is lost, please also refer to the schematic diagram of the process when the UAV makes a safe landing as shown in Figure 8, when the UAV determines that the navigation signal is lost , Can trigger the binocular vision module below and perform a visual return based on the position information provided by the visual odometer.
  • the drone can select any one of the at least one preset return position as the unmanned The target return position of the aircraft, assuming that the selected target return position is the starting position of the drone, that is, the position identified by point A in FIG. 7 and before the navigation signal is lost, the drone recorded towards the return target
  • the direction of the position is regarded as the first return direction.
  • the first return direction is the direction indicated by the black curve in FIG. 7, and the first return path can be further determined according to the first return direction and the first return target position (That is, the path indicated by the black curve), where the first return direction is the direction from the return home position to the return home target position, so that the drone can be controlled from the return home according to the indication of the first return home path Return from the starting position B to the returning target position A.
  • the visual odometer determines that the current position of the drone reaches the first preset range from the return home position, that is, when the drone is at point C shown in FIG. 7, in order to avoid the drone
  • Repeating the process of detecting the safety landing point can control the UAV to start the safety landing point detection and record the detected safety landing point.
  • the detected and recorded safe landing points are points a, b, and c marked with stars in FIG. 7.
  • the drone may fly to the point D in the second preset range from the return target position according to the instruction of the first return path to start a safe landing attempt; the drone may also Return to point A correctly and make a safe landing attempt at point A.
  • the safe landing attempt at point A or D fails, find the safe landing points (ie points a, b, and c) recorded in the second preset range, and determine from the recorded safe landing points
  • the safe landing point closest to point A or point D is point a, so after adjusting the flying height of the aircraft to the preset flying height, the first return path of the aircraft is adjusted to the second return path, so that The aircraft flies from point D to point a according to the indication of the second return path and performs a safe landing.
  • the aircraft does not find a recorded safe landing point, and fails to make multiple attempts to land when returning to the return target position, or if no safe landing point is detected within the second preset range, it can perform a hover-in-place to wait for the user Managed operation, and alarm to perform forced landing when severely low battery.
  • the drone hovering at the home position of the home may be the actual home position of the home, that is, point A, or it may be hovered at point D due to visual home error.
  • FIG. 9 is a structural diagram of a safety landing device applied to a drone according to an embodiment of the present invention.
  • the application The unmanned aerial vehicle landing device 900 includes a memory 901 and a processor 902, where the program code is stored in the memory 902, and the processor 902 calls the program code in the memory.
  • the processor 902 executes the following operations:
  • processor 902 is further configured to perform the following operations:
  • a safe landing point detection is performed, wherein the first preset range is greater than the second preset range ;
  • the step of landing according to the recorded safe landing point is performed;
  • the processor 902 performs the following operations when landing according to the recorded safe landing point:
  • the drone When the current position of the drone is the recorded safe landing point, the drone is controlled to land.
  • the processor 902 performs the following operations when landing according to the detected safe landing point:
  • the current flying height of the drone is adjusted to a preset flying height, and the step of landing according to the recorded safe landing point is performed;
  • the preset flying height is the flying height when the drone returns to flight.
  • processor 902 is further configured to perform the following operations:
  • the home return Perform a safe landing point detection according to a preset trajectory within a third preset range of the target position, where the preset trajectory includes a spiral trajectory or a polyline trajectory;
  • processor 902 is further configured to perform the following operations:
  • processor 902 is further configured to perform the following operations:
  • the detection result is that the home return target position is a safe landing point, landing is performed at the home return target position;
  • the landing step is performed according to the recorded safe landing point.
  • the processor 902 performs the following operations when determining the return home target position and the first return home path of the drone:
  • the first return path is determined according to the first return direction and the return target position.
  • the processor 902 performs the following operations when controlling the drone to return home based on the first return path and the return target location:
  • the second return direction corresponding to the second return path is opposite to the first return direction corresponding to the first return path.
  • the safe landing point is a plane position and a non-water surface position.
  • the processor 902 performs the following operations when detecting a safe landing point:
  • the processor 902 performs the following operations when detecting a safe landing point:
  • the navigation signal includes at least one of the following: a signal from a positioning sensor and a signal from a compass.
  • the processor 902 performs the following operations when performing plane detection on the current position of the drone according to a preset plane detection algorithm:
  • the processor 902 performs the following operations when performing plane detection on the current position according to the three-dimensional space point set:
  • the processor 902 performs the following operations when performing water surface detection on the current position of the drone according to a preset water surface detection algorithm:
  • the two-dimensional projection image is input to a convolutional neural network model, and it is determined whether the current position is a water surface according to the output of the convolutional neural network model.
  • the safety landing device applied to an unmanned aerial vehicle provided by this embodiment can execute the safety landing method shown in FIG. 1 and FIG. 2 provided in the foregoing embodiment, and the execution method and the beneficial effects are similar, and are not repeated here.
  • An embodiment of the present invention provides a drone, including a fuselage, a power system, and a safety landing device as described above.
  • the operation of the safety landing device of the UAV is the same as or similar to the foregoing, and will not be repeated here.
  • the power system of an unmanned aerial vehicle may include a rotor, a motor that drives the rotor to rotate, and its ESC.
  • the drone can be a four-rotor, six-rotor, eight-rotor, or other multi-rotor drone. It is understandable that the UAV can also be a fixed-wing UAV or a hybrid-wing UAV.
  • the drone provided by the embodiment of the present invention may further include a sensor installed on the fuselage.
  • the sensor includes a GNSS module for providing position information for the drone.
  • the sensor also includes at least one of a binocular vision sensor or a visual odometer.
  • a binocular vision sensor may be provided under the drone to obtain an image under the drone and generate a depth map, a semantic map, or other information, so as to detect a safe landing point.
  • the visual odometer may be set on the front side of the drone, so that the drone can provide flight mileage information for the drone when there is no GNSS signal or the GNSS module malfunctions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Multimedia (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Navigation (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

一种无人机的安全降落方法、装置、无人机及计算机存储介质，其中方法包括：当无人机丢失导航信号时，确定无人机的返航目标位置和第一返航路径（S101）；控制无人机基于第一返航路径和返航目标位置进行返航（S102）；当返航过程中无人机的当前位置在距离返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供无人机降落的安全降落点（S103）；根据记录的安全降落点进行降落（S104），该方法有助于为无人机的安全降落提供保障。

Description

一种无人机的安全降落方法、装置、无人机及介质 技术领域
本发明实施例涉及计算机技术领域，尤其涉及一种无人机的安全降落方法、装置、无人机及介质。
背景技术
农用无人机在自主作业的过程中，可能会由于环境干扰或者硬件故障等原因丢失GNSS信号导致农用无人机无法获得位置信息，为了使得无人机在丢失GNSS信号时能够实现返航，可借助机体下方的双目视觉模块，以使得农用无人机可根据该双目视觉模块提供的位置信息进行返航。
但是，由于光线、环境等原因使得双目视觉模块可能无法提供精准的位置，导致无人机在返航途中出现较大的偏差，由于该无人机返航时出现的较大偏差，可能使得无人机在返航途中降落到地面不平整的地区或者水中，使得无人机在降落时缺乏安全降落的保障。
发明内容
本发明实施例提供了一种无人机的安全降落方法、装置、无人机及介质，有助于为无人机的安全降落提供保障。
本发明实施例的第一方面提供了一种无人机的安全降落方法，该方法包括：
当所述无人机丢失导航信号时，确定无人机的返航目标位置和第一返航路径；
控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航；
当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点；
根据记录的安全降落点进行降落。
本发明实施例第二方面提供了一种安全降落装置，应用于无人机，其特征在于，所述安全降落装置包括存储器和处理器；
所述存储器用于存储程序代码；
所述处理器，调用所述程序代码，当程序代码被执行时，用于执行以下操 作：
当所述无人机丢失导航信号时，确定无人机的返航目标位置和第一返航路径；
控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航；
当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点；
根据记录的安全降落点进行降落。
本发明实施例的第三方面是提供的一种无人机，包括：
机身；
动力系统，安装在所述机身，用于为所述无人机提供动力；
以及如第二方面中所述的安全降落装置。
在本发明实施例中，当无人机丢失导航信号时，可在确定所述无人机的返航目标位置和第一返航路径后，控制所述无人机按照第一返航路径和所述返航目标位置进行返航，在检测到所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，检测并记录安全降落点，使得所述无人机可基于记录的所述安全降落点进行降落，可为所述无人机的安全降落提供可靠的保障，并同时提高了无人机在确定安全降落点时的工作效率，有效节省了所述无人机的处理资源。
附图说明
为了更清楚地说明本发明实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本发明实施例提供的一种无人机的安全降落方法的流程示意图；
图2是本发明另一实施例提供的一种无人机的安全降落方法的流程示意图；
图3是本发明实施例提供的一种无人机的观测区域和待检测区域的示意图；
图4是本发明实施例提供的一种将如图3所示的观测区域和待检测区域反投影成二维图像的示意图；
图5是本发明实施例提供的一种无人机的倾斜飞行姿态的示意图；
图6是本发明实施例提供的一种无人机按照预设轨迹确定安全降落点的示意图；
图7是本发明又一实施例提供的一种无人机的安全降落方法的流程示意图；
图8是本发明实施例提供的一种无人机的返航流程示意图；
图9是本发明实施例提供的一种安全降落装置的示意性框图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有付出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
无人机在作业过程中可能由于环境的干扰，或者硬件故障等因素导致无人机丢失导航信号。无人机的导航信号可以包括定位传感器的信号或指南针的信号中的至少一种。定位传感器可以包括无人机上设置的全球卫星导航系统(Global Navigation Satellite System，GNSS)模块。例如，进行农药喷洒作业的农用无人机可能由于环境因素对GNSS模块的干扰而发生故障(俗称为所述无人机丢星)，或者因为硬件故障而导致指南针失灵，从而造成无人机丢失导航信号。在所述无人机丢失导航信号时，无人机不能依靠GNSS模块或指南针获取到可靠的导航信号，也就是说此时的无人机不能根据可靠的坐标信息或方向信息执行精准返航，为了使丢失导航信号的无人机返航到返航目标位置或者该返航目标位置的附近位置，可借助无人机的双目视觉模块执行视觉返航，在一些实施例中，可将返航目标位置称为Home点。Home点可以是无人机的出发位置，也可以是用户自行设定的位置(可以不与出发位置相同)。当无人机执行返航时会将Home点作为返航目标位置并向其飞去。
当前，在基于双目视觉模块进行返航时，所述无人机可在丢失导航信号时， 可将该无人机的出发位置作为该无人机的返航目标位置，从而可根据无人机在丢失导航信号前最晚记录的所述无人机朝着返航目标位置的方向确定所述无人机的第一返航方向，并根据该第一返航方向和该返航目标位置，确定该无人机的第一返航路径，进一步地，可在所述无人机检测到丢失导航信号时根据所述第一返航路径的指示，向所述返航目标位置进行返航，当所述无人机返航到所述返航目标位置时，进行安全降落点检测，并在检测到安全降落点时执行降落操作；如果未检测到安全降落点，则需要重新寻找安全降落点并执行降落。
采用当前的无人机返航方法，虽然在到达返航目标位置时对当前的降落位置进行了安全降落点检测，但是如果检测结果指示当前的降落位置不是安全降落点，则需要无人机重新确定新的飞行路线以及重新从所述新的飞行路线中确定出安全降落点的步骤，降低了无人机确定安全降落点时的工作效率，浪费无人机的处理资源。
基于此，本申请提出了一种无人机的安全降落方法，可提高无人机在确定安全降落点时的工作效率，从而有效节省所述无人机的处理资源，并同时为所述无人机的安全降落提供可靠的保障。
请参见图1，是本发明实施例提供的一种无人机的安全降落方法的示意流程图，如图1所示，该方法可包括：
S101，当所述无人机丢失导航信号时，确定所述无人机的返航目标位置和第一返航路径。
在一个实施例中，无人机在飞行过程中如果发生定位故障，可能导致无人机丢失导航信号，例如在无人机的GNSS故障时导致的导航信号丢失情况。在所述无人机检测到导航信号丢失时，可触发无人机下方的双目视觉模块进行视觉返航，以使得所述无人机能返航到预设的返航目标位置处。在一个实施例中，所述预设的返航目标位置是从预设的至少一个返航位置中确定的，所述预设的返航位置例如可以是在所述无人机中预设的用于进行电池更换或者进行飞行作业的位置，也可以是无人机的出发位置。如果所述无人机仅由于丢失定位传感器的信号而导致导航信号的丢失，例如定位传感器发生故障而丢失导航信号，如上述的GNSS故障发生的导航信号的丢失，可依靠指南针确定无人机进行返航时的第一返航方向，从而结合确定出的返航目标位置以及该第一返航方向， 确定该无人机的第一返航路径。
在一个实施例中，无人机在飞行过程中如果发生指南针故障，也可能导致无人机丢失导航信号。此时无人机可以根据双目视觉模块进行视觉返航，并且在返航途中依据定位传感器的坐标信号确定自身的方向。所述预设的返航目标位置可以是如前所述的返航位置。此时无人机可以依靠定位传感器如GNSS模块确定无人机的位置，并结合确定出的返航目标位置确定该无人机的第一返航路径。
在另一实施例中，如果该无人机是由于丢失定位传感器和指南针的信号而导致的导航信号的丢失，如定位传感器和指南针均故障时导致的导航信号的丢失，无人机可以根据视觉模块对周围环境进行识别并据此定位自身的位置及方向，并且可以结合丢失导航信号前所记录的位置及方向信息来确定出该无人机的第一返航路径以及返航目标位置。
在无人机从返航目标位置飞行到丢失导航信号对应位置的过程中，无人机的定位装置正常，会对无人机的飞行位置进行实时的更新与记录。当无人机起飞时，还可记录所述无人机的返航目标位置。因此，当无人机在正常飞行过程中，正常工作的定位装置可以实时更新与记录无人机的当前位置，并且根据当前位置和返航目标位置，实时更新与记录无人机朝向返航目标位置的方向。
例如，无人机可以通过其上的惯性测量单元(Inertial Measurement Unit,IMU)实时记录无人机的姿态，从而实时记录无人机的飞行方向而获取无人机朝向返航目标位置的方向。当无人机丢失导航信号时，例如GNSS模块遇到干扰失去信号、故障或其他问题时，可从预设的至少一个返航位置中选取任一位置作为返航目标位置，并可将其丢失导航信号前最后记录的飞行位置确定为返航起始位置，以及将其丢失导航信号前最后记录的朝向返航目标位置的方向确定为第一返航方向，从而可根据该第一返航方向、返航目标位置以及返航起始位置，确定该无人机的第一返航路径。其中，所述第一返航方向为所述无人机的返航起始位置指向所述返航目标位置的方向。例如，无人机可以根据返航起始位置和第一返航方向，确定一条直线路径作为第一返航路径朝向Home点飞去；也可以根据返航起始位置和Home点，以及途中记录的飞行位置确定出原始飞行的路径并将其作为第一返航路径，返回至Home点；当然无人机也可以 根据其他方式生成第一返航路径，此处并不作限制。
S102，控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航。
在一个实施例中，无人机可将如GNSS模块等定位装置中记录的任一返航位置作为所述无人机的返航目标位置，所述GNSS模块等定位装置记录的任一返航位置例如可以是无人机进行飞行作业或者电池更换的位置，或者是农用无人机进行农药喷洒的位置，或者是用户所设定的位置等。在一个实施例中，无人机可结合所述无人机在返航过程中的当前位置实时调整所述无人机的第一返航路径，并控制所述无人机按照在所述返航过程中各当前位置对应的第一返航路径的指示朝着Home点(即返航目标位置)进行返航。
在一些情况下，第一返航路径可以是基本不变化的，例如无人机从返航起始位置直接朝向返航目标位置飞去。在另一些情况下，第一返航路径可以是变化的，例如无人机在返航过程中遇到障碍物并因避障功能启动而导致第一返航路径变化，此时无人机可以在绕开障碍物结束避障功能后，重新回到开始的第一返航路径，或者以变化后的第一返航路径继续返航。
在控制所述无人机基于所述第一返航路径和返航目标位置进行返航时，还可参考视觉里程计提供的位置信息进行，所述无人机可根据所述视觉里程计提供的位置信息确定所述无人机的当前位置。
S103，当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点。
在无人机丢失导航信号进行视觉返航的过程中，由于不能通过GNSS获取到所述无人机的准确坐标信息，再加上返航过程中视觉里程计和/或指南针的误差，可能使无人机不能精准地返航到返航目标位置，所述无人机最终的返航目标位置可能是所述无人机返航目标位置的附近位置，由于所述返航目标位置的附近位置可能不是适合无人机安全降落的位置，因此无人机需要重新确定出可供所述无人机降落的安全降落点，为了有效提高无人机寻找安全降落点的处理效率，可在返航过程中对所述无人机的当前位置进行安全降落点检测，并记录确定出的安全降落点，从而可在步骤S104中根据记录的安全降落点降落。
在一个实施例中，所述无人机可在当前位置距离所述返航目标位置的第一预设范围时，开始进行安全降落点检测，并将检测到的安全降落点进行记录(即存储检测到的安全降落点)，因此当无人机在Home点(即返航目标位置)或者Home点附近进行降落时，如果检测到该点为不安全的降落点，可执行步骤S104根据记录的安全降落点进行降落，避免了重复寻找安全降落点的情况，从而实现了对无人机处理资源的节省，可提高无人机降落到安全降落点的成功率。例如，无人机可以在当前位置距离所述返航目标位置30米范围时，即开始进行安全降落点检测，并记录所检测到的安全降落点，但此时并不降落而仍然飞向Home点。
S104，根据记录的所述安全降落点进行降落。
在一个实施例中，当所述无人机飞行到返航目标位置或者所述返航目标位置的附近位置时，可对其所处的当前位置进行安全降落点检测，如果检测到所述无人机所处的当前位置为安全降落点，则直接进行降落，并在降落过程中对该位置持续进行安全降落点检测，如果在降落过程中检测到该当前位置为不安全降落点，则调整飞行高度为执行降落之前的高度，并飞行到记录的安全降落点处进行降落；如果在降落过程中检测到该当前位置为安全降落点，则持续降低飞行高度，直到检测到所述无人机和地面之间的相对高度为零时关闭动力系统，使所述无人机降落到地面。
再一个实施例中，如果检测到所述无人机所处的当前位置不是安全降落点，则根据在距离所述返航目标位置的第一预设范围内记录的安全降落点进行降落，在所述无人机对所述安全降落点进行记录时，记录有所述安全降落点的位置信息，因此，当所述无人机向记录的所述安全降落点进行飞行时，可根据所述无人机的当前位置和记录的安全降落点的位置信息，确定所述无人机的第二返航路径，所述第二返航路径为无人机从当前位置飞向记录的安全记录点的路径。在一些情况下，所述无人机飞向记录的所述安全降落点的第二返航路径对应的第二返航方向与上述的第一返航路径对应的第一返航方向相反，因此，在所述无人机向所述记录的安全降落点飞行时，可直接将和所述第一返航方向相反的方向作为第二返航方向。
在本发明实施例中，当无人机丢失导航信号时，可在确定所述无人机的返 航目标位置和第一返航路径后，控制所述无人机按照第一返航路径和所述返航目标位置进行返航，在检测到所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，检测并记录安全降落点，使得所述无人机可基于记录的所述安全降落点进行降落，可为所述无人机的安全降落提供可靠的保障，并同时提高了无人机在确定安全降落点时的工作效率，有效节省了所述无人机的处理资源。
下面对本发明实施例中无人机对当前位置进行安全降落点检测的方法进行具体描述，如图2所示，包括如下步骤：
S201，当所述无人机丢失导航信号时，确定所述无人机的返航目标位置和第一返航路径。
S202，控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航。
在一个实施例中，步骤S201和步骤S202的具体实施方式可参见上述步骤S101和步骤S102的叙述，在此不再赘述。
S203，当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，按照预设的平面检测算法对所述无人机的当前位置进行平面检测，以及按照预设的水面检测算法对所述无人机的当前位置进行水面检测。
S204，记录可供所述无人机降落的安全降落点。
在步骤S203和步骤S204中，所述无人机可基于双目视觉传感器对所述无人机的当前位置进行平面检测，和/或对所述当前位置进行水面检测。所述无人机按照预设的平面检测算法对所述无人机的当前位置进行平面检测时，先从所述当前位置对应的观测区域中确定出进行平面检测的待检测区域，在一个实施例中，无人机的下视双目视觉传感器(即双目视觉模块)的观测区域较大，如图3所示的301区域(即由P 1-3D，P 2-3D，P 3-3D，P 4-3D围成的区域)假设为所述无人机双目视觉传感器的观测区域，而对所述当前位置进行平面检测时，只需要确定所述无人机在物理空间中对应的最大外接圆是否平整即可，不需要检测该双目视觉传感器可观测到的所有区域是否平整。
如果所述无人机选取的待检测区域太大，会导致无人机对当前位置的平面 检测结果出现误差，即如果所述无人机在当前位置对应的最大外接圆本应为平整地面，由于选取的待检测区域扩大后包含了不平整的地面，使得无人机认为当前位置不是平面，而出现安全降落点检测的误差；而如果所述无人机选取的待检测区域太小，也会导致无人机对当前位置的平面检测结果出现差错，即如果所述无人机所在的当前位置为不平整地面，而由于选取的待检测区域过小，使得选取的待检测区域检测为平整地面，则可能导致无人机将本不平整的当前位置误作为平整的当前位置，而控制无人机下降到该不平整的地面上而出现安全故障。
因此，在进行平面检测时，需要从观测区域301中选取出适当大小的待检测区域302进行平面检测，在一个实施例中，选取出的待检测区域302可以是物理空间中2米*2米的正方形区域，从而根据该平面检测结果确定当前位置是否为平面，在一个实施例中，所述观测区域301也可称作所述无人机的检测范围，所述待检测区域302可以称作所述无人机的图像兴趣范围(Region Of Interest，ROI)。可以理解的是，选取出的待检测区域302的区域尺寸并不限于2米*2米，在一些情况下，可以根据无人机的大小进行确定，例如，以无人机飞行时在地面的投影面尺寸作为待检测区域302的区域尺寸，或者将该投影面尺寸扩大50％作为区域尺寸。
在从所述无人机的观测区域中确定出进行平面检测的待检测区域后，进一步地，可确定所述待检测区域对应的二维投影图像，具体地，可将如图3所示的三维空间中的观测区域301以及待检测区域302通过反投影规则，得到二维图像中的观测区域图像和待检测区域图像，如图4所示，三维空间中的所述观测区域301对应于二维图像中的观测区域图像401(即由P 1-2D，P 2-2D，P 3-2D，P 4-2D围成的区域)，所述待检测区域302对应于二维图像中的待检测区域图像402。在一个实施例中，所述反投影规则具体可以是：
S·P 2d＝K·(R ci·P 3d+T ci)       式2.1
其中，以及P 3d为三维空间中待检测区域对应的三维空间点，P 2d为二维空间中和P 3d对应的二维图像点，基于相机的外参(R，T)和相机内参K，可将三维空间中待检测区域的各三维空间点按照上述式2.1对应的反投影规则进行反投影，可确定出待检测区域对应的二维投影图像402，在所述二维投影图像 402中的点即为需要处理观测的点。
在确定所述待检测区域对应的二维投影图像后，需将所述二维投影图像中的点转化为地面坐标系下的三维空间点，所述地面坐标系下的三维空间点为真实的观测点，用凸起点的凸起高度表示实际空间中高于平面的物体高度，例如，如果无人机下方选取的待检测区域为平面，得到的点云应该几乎集中在同一平面上；如果无人机下方选取的待检测区域包括一棵树，那么点云的形状包括类似树顶的凸起。
在一个实施例中，可采用双目视觉模块将所述二维投影图像中的点转化为地面坐标系下的三维空间点，即可使用双目视觉的深度图，将飞行器下方的待检测区域的物理场景转换为三维空间点集合，从而可根据所述三维空间点集合，对当前位置进行平面检测。
在根据所述三维空间点集合，对所述当前位置进行平面检测时，具体有两种实施方案：
(1)、使用平面方程拟合出和所述三维空间点集合最相近的拟合平面，然后判断所述拟合平面对于所述三维空间点集合的内点百分比，如果所述内点百分比满足预设的百分比数量，则将所述拟合平面确定为一个平面，并通过确定所述拟合平面的法向量确定所述拟合平面的倾斜程度。其中，所述平面方程为：
ax+by+cz+d＝0         式2.2
在一个实施例中，如式2.2所示的平面方程参数的拟合类似于求解线性系统Ax＝0，三维空间点集合中大量的点[x,y,z]和需要求解的参数a、b、c和d构成如式2.3所示的超定方程，从而可使用随机采样一致性算法(RANSCA)求解得到该平面方程。
Figure PCTCN2018117820-appb-000001
(2)、由于对所述当前位置进行平面检测即是需要确定当前位置对应的待检测区域是否为安全平面，且确定为安全的平面接近于水平面，因此，可根据公式cz+d＝0，将所要拟合的平面强制确定为水平面，则可通过判断所述三维 空间点集合关于所述强制拟合的水平面的内点百分比，并根据所述内点百分比确定当前位置是否平整。
在第二种实施方案中，无人机可先获取标准平面方程，并计算所述三维空间点集合中任一空间三维点与所述标准平面方程之间的距离，并根据所述距离确定所述三维空间点集合中的内点数量，所述内点为所述距离小于或等于预设距离阈值的三维空间点，并当所述内点数量大于或等于预设数量阈值时确定所述当前位置为平面。
在采用第二种方法确定当前位置是否平整时，由于强制拟合的水平面方程为cz+d＝0，因此，可令超定方程式2.3中的参数a＝b＝0，再对该超定方程进行求解，使得对超定方程的参数求解过程更为简单，在一个实施例中，根据超定方程求取的平面即为三维空间点集合中所有点云所聚集的平面，如果当前位置平整，则该点云所聚集的平面一定会拟合成一个水平面，如果当前位置不平整，则该点云所聚集的平面一定不能拟合成一个水平面，也就是说不能求解得到使得超定方程2.3有意义的参数。
在对超定方程式2.3求解得到平面方程后，根据所述三维空间点集合中各空间点与所述平面方程之间的距离，可将所述三维空间点集合中的各个空间点进行分类成内点和外点。在一个实施例中，可选取所述三维空间点集合中的任一空间点作为观测点，并计算所述观测点与所述平面方程之间的距离d，当d小于或等于预设的距离阈值时，将所述观测点分类为内点；当所述d大于所述预设的距离阈值时，将所述观测点分类为外点。其中，如果观测点坐标为(x、y、z)，则可按照式2.4计算所述观测点与所述平面方程之间的距离：
Figure PCTCN2018117820-appb-000002
需要说明的是，如果无人机以如图5所示的飞行姿态进行飞行，并在以所述飞行姿态进行飞行的过程中进行安全点检测，由于所述无人机的飞行姿态和水平面之间存在倾斜夹角，因此，在采用上述方法对当前位置进行平面检测时，需要根据所述无人机的飞行姿态、当前位置、倾斜角度以及飞行高度H对所述无人机的飞行高度进行补偿，并根据补偿后的飞行高度对所述无人机的当前位置进行平面检测。
再一个实施例中，在按照预设的水面检测算法对所述无人机的当前位置进 行水面检测时，也可根据所述无人机的双目视觉传感器从所述当前位置对应的观测区域301中确定出进行水面检测的待检测区域302，并按照相机的反投影规则确定所述待检测区域302对应的二维投影图像，如图4中的402区域。用于进行水面检测的机器学习算法，是基于灰度图训练得到的卷积神经网络CNN模型进行的，即可将得到的待检测区域对应的二维投影图像输入所述CNN模型，所述CNN模型可检测所述二维投影图像中每一帧图像对应的区域是否有水面存在，并将检测结果作为输出，以确定所述当前位置是否为水面。
在一个实施例中，根据上述的平面检测结果和水面检测结果，确定所述当前位置是否为安全降落点，并将确定的安全降落点进行记录，在记录所述安全降落点时，可记录所述安全降落点对应于返航目标位置的相对位置，其中，确定的所述安全降落点为当前位置是平面或者非水面的位置。
在一个实施例中，也可采用其他算法对当前位置进行平面检测和水面检测，例如，采用传统的基于视觉的检测方法进行平面检测和/或进行水面检测。
S205，当返航过程中所述无人机的当前位置在距离所述返航目标位置的第二预设范围时，进行安全降落点检测。
S206，若检测结果为未检测到安全降落点，则执行所述根据记录的安全降落点进行降落的步骤。
S207，若所述检测结果为检测到安全降落点，则根据检测到的安全降落点进行降落。
在步骤S205-步骤S207中，当所述无人机在距离所述返航目标位置的第一预设范围内时，可对当前位置进行安全降落点检测并记录下可供所述无人机降落的安全降落点，无人机在继续飞行到距离所述返航目标位置的第二预设范围时，所述当前位置距离所述返航目标位置的第二预设范围即为所述返航目标位置的附近，其中，所述第一预设范围大于所述第二预设范围，所述第一预设范围例如可以是20米或者10米等，所述第二预设范围例如可以是5米或者10米等。
当所述无人机的当前位置在距离所述返航目标位置的第二预设范围时，如果在进行安全降落点检测后的检测结果为未检测到安全降落点，则执行步骤S206根据记录的安全降落点进行降落的步骤。
在根据记录的安全降落点进行降落时，可先根据所述无人机的当前位置和记录的安全降落点确定第二返航路径，在根据所述当前位置和所述记录的安全降落点确定第二返航路径时，可所述第二返航路径为由所述当前位置(即在所述返航目标位置附近的位置)指向所述记录的安全降落点的方向，进一步地可基于所述第二返航路径，向记录的安全降落点飞行，并在所述双目视觉传感器确定所述无人机的当前位置为记录的所述安全降落点时，控制所述无人机进行降落。
在一个实施例中，如果在进行安全降落点检测后的检测结果为检测到安全降落点，则执行步骤S207根据检测到的安全降落点进行降落的步骤，具体地，无人机在降落过程中，需要持续检测所述检测的安全降落点是否安全，因为所述无人机在距离所述返航目标位置的第二预设范围的位置，通常是用户更换电池或者路边等位置，所以无人机在检测到当前位置为安全降落点并进行降落的过程中很可能有人出现，如果在下降过程中不持续检测所述安全降落点是否安全，也会出现无人机降落在不安全位置而引发安全故障的情况。
当所述无人机持续检测到所述检测的安全降落点安全时，成功降落到所述检测到的安全降落点；或者，如果在降落过程中检测到所述安全降落点不安全，则将所述无人机的当前飞行高度调整为预设飞行高度，所述预设飞行高度为所述无人机返航时的飞行高度(即执行降落之前的高度)，并执行上述S206中根据记录的安全降落点进行降落的步骤。
其中，所述预设飞行高度为保证所述无人机下方的双目视觉模块在返航过程中处于较高精度范围而设置的，所述预设飞行高度例如可以是2米或者2.5米等。
再一个实施例中，当所述无人机在返航过程中处于距离所述返航目标位置的第二预设范围内时未检测到安全降落点，并确定在第一预设范围内未记录有安全降落点时，可在距离所述返航目标位置的第三预设范围内按照预设轨迹进行安全降落点检测，其中，所述预设轨迹包括螺旋形轨迹、折线形轨迹或者直线型轨迹中的一种或多种，所述折线形轨迹例如可以是“Z”字形轨迹，从而可在所述预设轨迹中检测到安全降落点时，根据检测到的安全降落点降落。
举例来说，如果所述无人机从如图6的B点(即返航起始位置)向A点 (即返航目标位置)返航，在到达返航目标位置附近(即所述无人机的当前位置在距离所述返航目标位置的第二预设范围)的D点时，对所述当前位置(即D点)进行安全降落点检测，如果确定D点为不安全降落点，且确定所述无人机在第一预设范围内未记录有安全降落点，则可按照预设轨迹，例如可以是如图所示的灰色螺旋形轨迹从D点开始寻找安全降落点，如果在所述螺旋形轨迹的K点处检测到安全降落点，则降落在所述K点。
再一个实施例中，当所述无人机在返航过程中在距离所述返航目标位置的第二预设范围内未检测到安全降落点，并确定在距离所述返航目标位置的第一预设范围内未记录有安全降落点，还可控制所述无人机在所述返航目标位置悬停，以等待用户的操作指令，在所述无人机出现低电量告警时，可强制所述无人机执行降落。
再一个实施例中，当所述无人机的当前位置和所述返航目标位置重合时，即所述无人机进行视觉返航到返航目标位置时，可对该返航目标位置进行安全降落点检测，如果检测到所述返航目标位置为安全降落点，则在所述返航目标位置执行降落，如果检测到所述返航目标位置不是安全降落点，则执行步骤S206中根据记录的安全降落点进行降落的步骤。
在本发明实施例中，如果无人机丢失导航信号，可在确定无人机的返航目标位置和第一返航路径后，控制该无人机基于第一返航路径和返航目标位置进行返航，并在所述无人机的当前位置在距离返航目标位置的第一预设范围内时，按照预设的平面检测算法以及预设的水面检测算法对当前位置进行平面检测以及水面检测，为所述无人机的返航及降落过程提供了安全保障，可有效降低无人机在降落后发生安全故障的可能性，由于记录了确定为平面的安全降落点，和/或非水面的安全降落点，无人机在返航到距离所述返航目标位置的第二预设范围时，可进行安全降落点检测，并进行安全降落尝试，如果所述无人机在距离所述返航目标位置的第二预设范围的当前位置没有检测到安全降落点，就根据记录的安全降落点进行降落，如果检测到安全降落点则直接进行降落，可提高无人机对安全降落点的确定速度，从而提高了无人机安全降落速度。
在本发明实施例中，提出了一种基于无人机的安全降落方法的应用场景图，如图7所示，无人机从A点起飞，并根据GNSS提供的导航信号向B飞行， 在图7中用灰色曲线标识所述无人机从A点飞向B点的飞行轨迹，在所述无人机从A点向B点飞行的过程中，会根据GNSS提供的可靠坐标信息实时刷新并记录所述无人机的位置信息。
如果所述无人机在B点时发生GNSS故障而丢失导航信号，请一并参考如图8所示的所述无人机进行安全降落时的流程示意图，无人机在确定丢失导航信号时，可触发下方的双目视觉模块并根据视觉里程计提供的位置信息进行视觉返航，在进行视觉返航时，无人机可从预设的至少一个返航位置中选取任一位置作为所述无人机的返航目标位置，假设选取的目标返航位置为该无人机的出发位置，即图7中A点标识的位置并将丢失该导航信号前所述无人机记录的朝着所述返航目标位置的方向，作为第一返航方向，所述第一返航方向即是图7中黑色曲线所指示的方向，进一步地可根据该第一返航方向和该第一返航目标位置，确定第一返航路径(即如图黑色曲线所指示的路径)，其中，该第一返航方向为由返航起始位置指向返航目标位置的方向，从而可根据该第一返航路径的指示，控制无人机从返航起始位置B点向返航目标位置A点进行返航。
当所述视觉里程计确定所述无人机的当前位置到达距离返航目标位置的第一预设范围时，即当所述无人机处于图7所示的C点时，为了避免无人机重复检测安全降落点的过程，可控制所述无人机开始进行安全降落点检测，并记录检测到的安全降落点。在一个实施例中，假设检测并记录的安全降落点为图7中用星标记的a、b和c点。
由于视觉返航导致的返航误差，所述无人机可能按照所述第一返航路径的指示飞行到距离返航目标位置第二预设范围的D点开始进行安全降落尝试；所述无人机也可能正确地回到A点，并在A点进行安全降落尝试。如果在A点或D点进行的安全降落尝试失败，则查找在第二预设范围内记录的安全降落点(即a、b和c点)，并从所述记录的安全降落点中确定出离A点或D点最近的安全降落点，即为a点，从而调整所述飞行器的飞行高度为预设的飞行高度后，将所述飞行器的第一返航路径调整为第二返航路径，使所述飞行器根据第二返航路径的指示从D点飞行到a点并执行安全降落。
如果所述飞行器没有查找到记录的安全降落点，且回到返航目标位置时多 次尝试降落失败，或者在第二预设范围内没有检测到安全降落点，则可执行原地悬停等待用户托管的操作，并在严重低电量时报警执行强制降落。可以理解的是，无人机在返航目标位置悬停，可以是实际返航目标位置即A点悬停，也可以由于视觉返航误差而导致在D点悬停。
本发明实施例提供了一种安全降落装置，应用于无人机中，图9是本发明实施例提供应用于无人机的安全降落装置的结构图，如图9所示，所述应用于无人机的安全降落装置900包括存储器901和处理器902，其中，存储器902中存储有程序代码，处理器902调用存储器中的程序代码，当程序代码被执行时，处理器902执行如下操作：
当所述无人机丢失导航信号时，确定所述无人机的返航目标位置和第一返航路径；
控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航；
当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点；
根据记录的安全降落点进行降落。
在一个实施例中，所述处理器902还用于执行如下操作：
当返航过程中所述无人机的当前位置在距离所述返航目标位置的第二预设范围时，进行安全降落点检测，其中，所述第一预设范围大于所述第二预设范围；
若检测结果为未检测到安全降落点，则执行所述的根据记录的安全降落点进行降落的步骤；
若所述检测结果为检测到安全降落点，则根据检测到的安全降落点进行降落。
在一个实施例中，所述处理器902在根据记录的安全降落点进行降落时，执行如下操作：
根据所述无人机的当前位置和记录的安全降落点，确定第二返航路径；
基于所述第二返航路径，向所述记录的安全降落点飞行；
当所述无人机的当前位置为所述记录的安全降落点时，控制所述无人机进 行降落。
在一个实施例中，所述处理器902在根据检测到的安全降落点进行降落时，执行如下操作：
在降落过程中，持续检测所述检测的安全降落点是否安全；
若持续检测到所述检测的安全降落点均安全，则确定成功降落到所述检测到的安全降落点；
若检测到所述检测的安全降落点不安全，则将无人机的当前飞行高度调整为预设飞行高度，并执行所述根据记录的安全降落点进行降落的步骤；
其中，所述预设飞行高度为所述无人机返航时的飞行高度。
在一个实施例中，所述处理器902还用于执行如下操作：
若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则在距离所述返航目标位置的第三预设范围内按照预设轨迹进行安全降落点检测，所述预设轨迹包括螺旋形轨迹或折线形轨迹；
在所述预设轨迹中检测到安全降落点时，根据检测到的安全降落点进行降落。
在一个实施例中，所述处理器902还用于执行如下操作：
若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则控制所述无人机在所述返航目标位置悬停。
在一个实施例中，所述处理器902还用于执行如下操作：
当所述无人机到达所述返航目标位置时，对所述返航目标位置进行安全降落点检测；
若检测结果为所述返航目标位置是安全降落点，则在所述返航目标位置执行降落；
若检测结果为所述返航目标位置不是安全降落点，则执行根据记录的安全降落点进行降落的步骤。
在一个实施例中，所述处理器902在确定所述无人机的返航目标位置和第一返航路径时，执行如下操作：
从预设的至少一个返航位置中选取任一位置作为所述无人机的返航目标位置；
根据所述返航目标位置确定第一返航方向；
根据所述第一返航方向和所述返航目标位置，确定所述第一返航路径。
在一个实施例中，所述处理器902在控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航时，执行如下操作：
控制所述无人机基于所述第一返航路径和所述返航目标位置，以及视觉里程计提供的位置信息进行返航。
在一个实施例中，所述第二返航路径对应的第二返航方向和所述第一返航路径对应的第一返航方向相反。
在一个实施例中，所述安全降落点为平面位置且非水面的位置。
在一个实施例中，所述处理器902在进行安全降落点检测时，执行如下操作：
基于双目视觉传感器进行安全降落点检测。
在一个实施例中，所述处理器902在进行安全降落点检测时，执行如下操作：
按照预设的平面检测算法对所述无人机的当前位置进行平面检测，以及按照预设的水面检测算法对所述无人机的当前位置进行水面检测。
在一个实施例中，所述导航信号包括如下至少一种：定位传感器的信号，指南针的信号。
在一个实施例中，所述处理器902在按照预设的平面检测算法对所述无人机的当前位置进行平面检测时，执行如下操作：
从所述当前位置对应的观测区域中确定出进行平面检测的待检测区域，所述待检测区域小于所述观测区域；
确定所述待检测区域对应的二维投影图像；
将所述二维投影图像中任一像素点转换为三维空间点，得到所述二维投影图像对应的三维空间点集合；
根据所述三维空间点集合，对所述当前位置进行平面检测。
在一个实施例中，所述处理器902在根据所述三维空间点集合，对所述当 前位置进行平面检测时，执行如下操作：
获取标准平面方程；
计算所述三维空间点集合中任一三维空间点与所述标准平面方程之间的距离，并根据所述距离确定所述三维空间点集合中的内点数量，所述内点为所述距离小于或等于预设距离阈值的三维空间点；
当所述内点数量大于或等于预设数量阈值时，确定所述当前位置为平面。
在一个实施例中，所述处理器902在按照预设的水面检测算法对所述无人机的当前位置进行水面检测时，执行如下操作：
从所述当前位置对应的观测区域中确定出进行水面检测的待检测区域，所述待检测区域小于所述观测区域；
确定所述待检测区域对应的二维投影图像；
将所述二维投影图像输入卷积神经网络模型，根据所述卷积神经网络模型的输出确定所述当前位置是否为水面。
本实施例提供的应用于无人机的安全降落装置能执行前述实施例提供的如图1和图2所示的安全降落方法，且执行方式和有益效果类似，在这里不再赘述。
本发明实施例提供一种无人机，包括机身，动力系统以及如前所述的安全降落装置。无人机的安全降落装置工作与前述相同或类似，此处不再赘述。无人机的动力系统可以包括旋翼、驱动旋翼旋转的电机及其电调。无人机可以是四旋翼、六旋翼、八旋翼或其他多旋翼无人机，此时无人机垂直起降进行工作。可以理解的是，无人机还可以是固定翼无人机或混合翼无人机。
本发明实施例提供的无人机还可以包括安装于机身上的传感器。传感器包括GNSS模块，用于为所述无人机提供位置信息。传感器还包括双目视觉传感器或者视觉里程计中的至少一种。在一些实施方式中，双目视觉传感器可以设置在无人机的下方，用于获取无人机下方的图像，并生成深度图、语义图或其他信息，从而进行安全降落点检测。在一些实施方式中，视觉里程计可以设置在无人机的前侧，从而使得无人机在没有GNSS信号或GNSS模块故障无法工作时，为无人机提供飞行里程信息。
最后应说明的是：以上各实施例仅用以说明本发明的技术方案，而非对其 限制；尽管参照前述各实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分或者全部技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的范围。

Claims (37)

  1. 一种无人机的安全降落方法，其特征在于，包括：
    当所述无人机丢失导航信号时，确定所述无人机的返航目标位置和第一返航路径；
    控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航；
    当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点；
    根据记录的安全降落点进行降落。
  2. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    当返航过程中所述无人机的当前位置在距离所述返航目标位置的第二预设范围时，进行安全降落点检测，其中，所述第一预设范围大于所述第二预设范围；
    若检测结果为未检测到安全降落点，则执行所述根据记录的安全降落点进行降落的步骤；
    若所述检测结果为检测到安全降落点，则根据检测到的安全降落点进行降落。
  3. 根据权利要求2所述的方法，其特征在于，所述根据记录的安全降落点进行降落，包括：
    根据所述无人机的当前位置和记录的安全降落点，确定第二返航路径；
    基于所述第二返航路径，向所述记录的安全降落点飞行；
    当所述无人机的当前位置为所述记录的安全降落点时，控制所述无人机进行降落。
  4. 根据权利要求2所述的方法，其特征在于，所述根据检测到的安全降落点进行降落，包括：
    在降落过程中，持续检测所述检测到的安全降落点是否安全；
    若持续检测到所述检测的安全降落点均安全，则确定成功降落到所述检测到的安全降落点；
    若检测到所述检测的安全降落点不安全，则将无人机的当前飞行高度调整为预设飞行高度，并执行所述根据记录的安全降落点进行降落的步骤；
    其中，所述预设飞行高度为所述无人机返航时的飞行高度。
  5. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则在距离所述返航目标位置的第三预设范围内按照预设轨迹进行安全降落点检测，所述预设轨迹包括螺旋形轨迹或折线形轨迹；
    在所述预设轨迹中检测到安全降落点时，根据检测到的安全降落点进行降落。
  6. 根据权利要求2所述的方法，其特征在于，所述方法还包括：
    若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则控制所述无人机在所述返航目标位置悬停。
  7. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    当所述无人机到达所述返航目标位置时，对所述返航目标位置进行安全降落点检测；
    若检测结果为所述返航目标位置是安全降落点，则在所述返航目标位置执行降落；
    若检测结果为所述返航目标位置不是安全降落点，则执行根据记录的安全降落点进行降落的步骤。
  8. 根据权利要求1所述的方法，其特征在于，所述确定所述无人机的返航目标位置和第一返航路径，包括：
    从预设的至少一个返航位置中选取任一位置作为所述无人机的返航目标位置；
    根据所述返航目标位置确定第一返航方向；
    根据所述第一返航方向和所述返航目标位置，确定所述第一返航路径。
  9. 根据权利要求1所述的方法，其特征在于，所述控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航，包括：
    控制所述无人机基于所述第一返航路径和所述返航目标位置，以及视觉里程计提供的位置信息进行返航。
  10. 根据权利要求3所述的方法，其特征在于，所述第二返航路径对应的第二返航方向和所述第一返航路径对应的第一返航方向相反。
  11. 根据权利要求1-10任一项所述的方法，其特征在于，所述安全降落点为平面且非水面的位置。
  12. 根据权利要求1-10任一项所述的方法，其特征在于，所述进行安全降落点检测，包括：
    基于双目视觉传感器进行安全降落点检测。
  13. 根据权利要求1-10任一项所述的方法，其特征在于，所述进行安全降落点检测，包括：
    按照预设的平面检测算法对所述无人机的当前位置进行平面检测，以及按照预设的水面检测算法对所述无人机的当前位置进行水面检测。
  14. 根据权利要求1-10任一项所述的方法，其特征在于，所述导航信号包括如下至少一种：定位传感器的信号，指南针的信号。
  15. 根据权利要求13所述的方法，其特征在于，所述按照预设的平面检 测算法对所述无人机的当前位置进行平面检测，包括:
    从所述当前位置对应的观测区域中确定出进行平面检测的待检测区域，所述待检测区域小于所述观测区域；
    确定所述待检测区域对应的二维投影图像；
    将所述二维投影图像中任一像素点转换为三维空间点，得到所述二维投影图像对应的三维空间点集合；
    根据所述三维空间点集合，对所述当前位置进行平面检测。
  16. 根据权利要求14所述的方法，其特征在于，所述根据所述三维空间点集合，对所述当前位置进行平面检测，包括：
    获取标准平面方程；
    计算所述三维空间点集合中任一三维空间点与所述标准平面方程之间的距离，并根据所述距离确定所述三维空间点集合中的内点数量，所述内点为所述距离小于或等于预设距离阈值的三维空间点；
    当所述内点数量大于或等于预设数量阈值时，确定所述当前位置为平面。
  17. 根据权利要求13所述的方法，其特征在于，所述按照预设的水面检测算法对所述无人机的当前位置进行水面检测，包括：
    从所述当前位置对应的观测区域中确定出进行水面检测的待检测区域，所述待检测区域小于所述观测区域；
    确定所述待检测区域对应的二维投影图像；
    将所述二维投影图像输入卷积神经网络模型，根据所述卷积神经网络模型的输出确定所述当前位置是否为水面。
  18. 一种安全降落装置，应用于无人机，其特征在于，所述安全降落装置包括存储器和处理器；
    所述存储器用于存储程序代码；
    所述处理器，调用所述程序代码，当程序代码被执行时，用于执行以下操作：
    当所述无人机丢失导航信号时，确定所述无人机的返航目标位置和第一返航路径；
    控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航；
    当返航过程中所述无人机的当前位置在距离所述返航目标位置的第一预设范围内时，进行安全降落点检测，并记录可供所述无人机降落的安全降落点；
    根据记录的安全降落点进行降落。
  19. 根据权利要求18所述的装置，其特征在于，所述装置还用于执行如下操作：
    当返航过程中所述无人机的当前位置在距离所述返航目标位置的第二预设范围时，进行安全降落点检测，其中，所述第一预设范围大于所述第二预设范围；
    若检测结果为未检测到安全降落点，则执行所述的根据记录的安全降落点进行降落的步骤；
    若所述检测结果为检测到安全降落点，则根据检测到的安全降落点进行降落。
  20. 根据权利要求19所述的装置，其特征在于，所述根据记录的安全降落点进行降落时，执行如下操作：
    根据所述无人机的当前位置和记录的安全降落点，确定第二返航路径；
    基于所述第二返航路径，向所述记录的安全降落点飞行；
    当所述无人机的当前位置为所述记录的安全降落点时，控制所述无人机进行降落。
  21. 根据权利要求19所述的装置，其特征在于，所述根据检测到的安全降落点进行降落时，执行如下操作：
    在降落过程中，持续检测所述检测的安全降落点是否安全；
    若持续检测到所述检测的安全降落点均安全，则确定成功降落到所述检测到的安全降落点；
    若检测到所述检测的安全降落点不安全，则将无人机的当前飞行高度调整为预设飞行高度，并执行所述根据记录的安全降落点进行降落的步骤；
    其中，所述预设飞行高度为所述无人机返航时的飞行高度。
  22. 根据权利要求19所述的装置，其特征在于，所述装置还用于执行如下操作：
    若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则在距离所述返航目标位置的第三预设范围内按照预设轨迹进行安全降落点检测，所述预设轨迹包括螺旋形轨迹或折线形轨迹；
    在所述预设轨迹中检测到安全降落点时，根据检测到的安全降落点进行降落。
  23. 根据权利要求19所述的装置，其特征在于，所述装置还用于执行如下操作：
    若在距离所述返航目标位置的第一预设范围内未记录有安全降落点，且在距离所述返航目标位置的第二预设范围内未检测到安全降落点，则控制所述无人机在所述返航目标位置悬停。
  24. 根据权利要求18所述的装置，其特征在于，所述装置还用于执行如下操作：
    当所述无人机到达所述返航目标位置时，对所述返航目标位置进行安全降落点检测；
    若检测结果为所述返航目标位置是安全降落点，则在所述返航目标位置执行降落；
    若检测结果为所述返航目标位置不是安全降落点，则执行根据记录的安全降落点进行降落的步骤。
  25. 根据权利要求18所述的装置，其特征在于，所述确定所述无人机的 返航目标位置和第一返航路径时，执行如下操作：
    从预设的至少一个返航位置中选取任一位置作为所述无人机的返航目标位置；
    根据所述返航目标位置确定第一返航方向；
    根据所述第一返航方向和所述返航目标位置，确定所述第一返航路径。
  26. 根据权利要求18所述的装置，其特征在于，所述控制所述无人机基于所述第一返航路径和所述返航目标位置进行返航时，执行如下操作：
    控制所述无人机基于所述第一返航路径和所述返航目标位置，以及视觉里程计提供的位置信息进行返航。
  27. 根据权利要求20所述的装置，其特征在于，所述第二返航路径对应的第二返航方向和所述第一返航路径对应的第一返航方向相反。
  28. 根据权利要求18-27任一项所述的装置，其特征在于，所述安全降落点为平面且非水面的位置。
  29. 根据权利要求18-27任一项所述的装置，其特征在于，所述进行安全降落点检测时，执行如下操作：
    基于双目视觉传感器进行安全降落点检测。
  30. 根据权利要求18-27任一项所述的装置，其特征在于，所述进行安全降落点检测时，执行如下操作：
    按照预设的平面检测算法对所述无人机的当前位置进行平面检测，以及按照预设的水面检测算法对所述无人机的当前位置进行水面检测。
  31. 根据权利要求18-27任一项所述的装置，其特征在于，所述导航信号包括如下至少一种：定位传感器的信号，指南针的信号。
  32. 根据权利要求30所述的装置，其特征在于，所述按照预设的平面检测算法对所述无人机的当前位置进行平面检测时，执行如下操作：
    从所述当前位置对应的观测区域中确定出进行平面检测的待检测区域，所述待检测区域小于所述观测区域；
    确定所述待检测区域对应的二维投影图像；
    将所述二维投影图像中任一像素点转换为三维空间点，得到所述二维投影图像对应的三维空间点集合；
    根据所述三维空间点集合，对所述当前位置进行平面检测。
  33. 根据权利要求31所述的装置，其特征在于，所述根据所述三维空间点集合，对所述当前位置进行平面检测时，执行如下操作：
    获取标准平面方程；
    计算所述三维空间点集合中任一三维空间点与所述标准平面方程之间的距离，并根据所述距离确定所述三维空间点集合中的内点数量，所述内点为所述距离小于或等于预设距离阈值的三维空间点；
    当所述内点数量大于或等于预设数量阈值时，确定所述当前位置为平面。
  34. 根据权利要求30所述的装置，其特征在于，所述按照预设的水面检测算法对所述无人机的当前位置进行水面检测时，执行如下操作：
    从所述当前位置对应的观测区域中确定出进行水面检测的待检测区域，所述待检测区域小于所述观测区域；
    确定所述待检测区域对应的二维投影图像；
    将所述二维投影图像输入卷积神经网络模型，根据所述卷积神经网络模型的输出确定所述当前位置是否为水面。
  35. 一种无人机，其特征在于，包括：
    机身；
    动力系统，安装在所述机身，用于为所述无人机提供动力；
    以及如权利要求18-34中任一项所述的安全降落装置。
  36. 根据权利要求35所述的无人机，其特征在于，所述无人机还包括：
    传感器，安装在所述机身，所述传感器至少包括如下一种：双目视觉传感器或者视觉里程计；
    其中，所述双目视觉传感器用于进行安全降落点检测；
    所述视觉里程计用于提供所述无人机返航时的位置信息。
  37. 一种计算机存储介质，其特征在于，所述计算机存储介质中存储有计算机程序指令，所述计算机程序指令被处理器执行时，用于执行如权利要求1-17任一项所述的无人机的安全降落方法。
PCT/CN2018/117820 2018-11-28 2018-11-28 一种无人机的安全降落方法、装置、无人机及介质 WO2020107248A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2018/117820 WO2020107248A1 (zh) 2018-11-28 2018-11-28 一种无人机的安全降落方法、装置、无人机及介质
CN201880066282.1A CN111615677B (zh) 2018-11-28 2018-11-28 一种无人机的安全降落方法、装置、无人机及介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2018/117820 WO2020107248A1 (zh) 2018-11-28 2018-11-28 一种无人机的安全降落方法、装置、无人机及介质

Publications (1)

Publication Number Publication Date
WO2020107248A1 true WO2020107248A1 (zh) 2020-06-04

Family

ID=70854244

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/117820 WO2020107248A1 (zh) 2018-11-28 2018-11-28 一种无人机的安全降落方法、装置、无人机及介质

Country Status (2)

Country Link
CN (1) CN111615677B (zh)
WO (1) WO2020107248A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114578855B (zh) * 2022-03-03 2022-09-20 北京新科汇智科技发展有限公司 一种无人机备降方法及系统

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104049641A (zh) * 2014-05-29 2014-09-17 深圳市大疆创新科技有限公司 一种自动降落方法、装置及飞行器
CN104881039A (zh) * 2015-05-12 2015-09-02 零度智控（北京）智能科技有限公司 一种无人机返航的方法及系统
US20160132057A1 (en) * 2013-07-09 2016-05-12 Duretek Inc. Method for constructing air-observed terrain data by using rotary wing structure
CN105867423A (zh) * 2016-06-08 2016-08-17 杨珊珊 无人飞行器返航方法、返航系统及其无人飞行器
US20170050747A1 (en) * 2015-08-22 2017-02-23 Olaf Wessler Method for destination approach control of unmanned aerial vehicles
CN107291099A (zh) * 2017-07-06 2017-10-24 杨顺伟 无人机返航方法及装置
CN107479082A (zh) * 2017-09-19 2017-12-15 广东容祺智能科技有限公司 一种无人机无gps返航方法
CN107943090A (zh) * 2017-12-25 2018-04-20 广州亿航智能技术有限公司 一种无人机的降落方法及系统
CN108124471A (zh) * 2017-12-11 2018-06-05 深圳市道通智能航空技术有限公司 无人飞行器返航方法、装置、存储介质和无人飞行器
CN108474658A (zh) * 2017-06-16 2018-08-31 深圳市大疆创新科技有限公司 地面形态检测方法及系统、无人机降落方法和无人机

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106527481A (zh) * 2016-12-06 2017-03-22 重庆零度智控智能科技有限公司 无人机飞行控制方法、装置及无人机
CN106927059A (zh) * 2017-04-01 2017-07-07 成都通甲优博科技有限责任公司 一种基于单目视觉的无人机降落方法及装置

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160132057A1 (en) * 2013-07-09 2016-05-12 Duretek Inc. Method for constructing air-observed terrain data by using rotary wing structure
CN104049641A (zh) * 2014-05-29 2014-09-17 深圳市大疆创新科技有限公司 一种自动降落方法、装置及飞行器
CN104881039A (zh) * 2015-05-12 2015-09-02 零度智控（北京）智能科技有限公司 一种无人机返航的方法及系统
US20170050747A1 (en) * 2015-08-22 2017-02-23 Olaf Wessler Method for destination approach control of unmanned aerial vehicles
CN105867423A (zh) * 2016-06-08 2016-08-17 杨珊珊 无人飞行器返航方法、返航系统及其无人飞行器
CN108474658A (zh) * 2017-06-16 2018-08-31 深圳市大疆创新科技有限公司 地面形态检测方法及系统、无人机降落方法和无人机
CN107291099A (zh) * 2017-07-06 2017-10-24 杨顺伟 无人机返航方法及装置
CN107479082A (zh) * 2017-09-19 2017-12-15 广东容祺智能科技有限公司 一种无人机无gps返航方法
CN108124471A (zh) * 2017-12-11 2018-06-05 深圳市道通智能航空技术有限公司 无人飞行器返航方法、装置、存储介质和无人飞行器
CN107943090A (zh) * 2017-12-25 2018-04-20 广州亿航智能技术有限公司 一种无人机的降落方法及系统

Also Published As

Publication number Publication date
CN111615677B (zh) 2024-04-12
CN111615677A (zh) 2020-09-01

Similar Documents

Publication Publication Date Title
Sani et al. Automatic navigation and landing of an indoor AR. drone quadrotor using ArUco marker and inertial sensors
US11242144B2 (en) Aerial vehicle smart landing
EP3903164B1 (en) Collision avoidance system, depth imaging system, vehicle, map generator, amd methods thereof
Zhang et al. 2d lidar-based slam and path planning for indoor rescue using mobile robots
WO2020103034A1 (zh) 一种无人机路径规划方法、装置及无人机
JP7263630B2 (ja) 無人航空機による３次元再構成の実行
EP3771956B1 (en) Systems and methods for generating flight paths for navigating an aircraft
WO2020181719A1 (zh) 无人机控制方法、无人机及系统
CN109407708A (zh) 一种基于多信息融合的精准着陆控制系统及着陆控制方法
US20220055748A1 (en) Obstacle avoidance method and apparatus for unmanned aerial vehicle landing, and unmanned aerial vehilce
US20220198793A1 (en) Target state estimation method and apparatus, and unmanned aerial vehicle
US11715072B2 (en) System, devices and methods for tele-operated robotics
WO2020014951A1 (zh) 建立局部障碍物地图的方法、装置及无人机
CN112789672B (zh) 控制和导航系统、姿态优化、映射和定位技术
EP3893078A1 (en) Relay point generation method and apparatus, and unmanned aerial vehicle
CN112596071A (zh) 无人机自主定位方法、装置及无人机
CN112379681A (zh) 无人机避障飞行方法、装置及无人机
US20210200246A1 (en) Method and system for determining the position of a moving object
CN112378397A (zh) 无人机跟踪目标的方法、装置及无人机
KR20210075647A (ko) 깊이 카메라를 이용한 무인 비행체의 비행 제어를 위한 학습 방법 및 장치
WO2020107248A1 (zh) 一种无人机的安全降落方法、装置、无人机及介质
US20230095700A1 (en) Vehicle flight control method and apparatus for unmanned aerial vehicle, and unmanned aerial vehicle
CN112380933A (zh) 无人机识别目标的方法、装置及无人机
KR102249485B1 (ko) 로봇 자율주행 시스템 및 방법
d'Apolito et al. Obstacle avoidance system development for the Ardrone 2.0 using the tum_ardrone package

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18941648

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18941648

Country of ref document: EP

Kind code of ref document: A1