CN110456805B - Intelligent tracking flight system and method for unmanned aerial vehicle - Google Patents
Intelligent tracking flight system and method for unmanned aerial vehicle Download PDFInfo
- Publication number
- CN110456805B CN110456805B CN201910548062.9A CN201910548062A CN110456805B CN 110456805 B CN110456805 B CN 110456805B CN 201910548062 A CN201910548062 A CN 201910548062A CN 110456805 B CN110456805 B CN 110456805B
- Authority
- CN
- China
- Prior art keywords
- flight
- unmanned aerial
- aerial vehicle
- starting
- detection
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
Images
Classifications
-
- 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
- G05D1/08—Control of attitude, i.e. control of roll, pitch, or yaw
- G05D1/0808—Control of attitude, i.e. control of roll, pitch, or yaw specially adapted for aircraft
-
- 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
- G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
Abstract
The invention relates to the field of intelligent autonomous navigation of unmanned aerial vehicles, in particular to an intelligent tracking flight system and method of an unmanned aerial vehicle. In the flight process, adopt the flight attitude measuring unit based on the tunnel magnetoresistance effect of high sensitive characteristic, through flight attitude measuring unit acquires unmanned aerial vehicle at the change key position of flight attitude, and starts the image acquisition unit carries out the focus and catches the image, forms the three-dimensional model of flight environment, provides key position image for unmanned aerial vehicle tracking flight route, improves the whole security of unmanned aerial vehicle tracking flight.
Description
Technical Field
The invention relates to the field of intelligent autonomous navigation of unmanned aerial vehicles, in particular to an intelligent tracking flight system and method of an unmanned aerial vehicle.
Background
At present, the core research problem of the intelligent unmanned aerial vehicle is the research of positioning and obstacle avoidance. And keep away barrier function and be one of the indispensable key index of weighing intelligent unmanned aerial vehicle intellectuality. If realize that intelligent unmanned aerial vehicle walking in-process accomplishes independently keeping away the barrier, the problem that will solve is: on the one hand, external environment information where the robot is located is required, and on the other hand, it is required that acquired environment information can be converted into control information through appropriate processing. The camera on the unmanned aerial vehicle can gather the image in unmanned aerial vehicle the place ahead, then can discern this image and send control command based on convolutional neural network's image identification module, and then help unmanned aerial vehicle to keep on the exact route, realizes the autonomic tracking flight of unmanned aerial vehicle in the corridor.
In recent years, convolutional neural networks have made a great breakthrough in image and speech recognition, and have become a research hotspot in the field of current speech analysis and image recognition. CNN is proposed as a deep learning architecture to minimize the preprocessing requirements of the data. However, the unmanned aerial vehicle on the market at present does not have the problem of good contact with the tracking method.
Further, in the field of unmanned aerial vehicles, the tunnel magnetoresistance effect is based on the spin effect of electrons, and has the advantages of high sensitivity, miniaturization and easiness in detection, but the tunnel magnetoresistance effect is not applied to the angular rate detection of the unmanned aerial vehicle at present, and the micro gyroscope for detecting the tunnel magnetoresistance effect in actual work faces the difficult problems that a stable detection magnetic field cannot be provided, reasonable assembly cannot be achieved and the like.
Disclosure of Invention
In order to effectively solve the problems, the invention provides an unmanned aerial vehicle intelligent tracking flight system and method based on a toroidal tunnel magnetoresistive micro-gyroscope.
The specific technical scheme of the invention is as follows:
an intelligent tracking flight system of an unmanned aerial vehicle is applied to the flight of the unmanned aerial vehicle, and comprises an image acquisition unit for acquiring images, a flight attitude measurement unit for detecting flight inertia, a neural network processing unit for processing neural network and a control unit for controlling the flight action of the unmanned aerial vehicle;
the image acquisition unit is connected with the neural network processing unit to image data transmission who will gather extremely among the neural network processing unit, flight attitude measurement unit real-time detection unmanned aerial vehicle's flight gesture to when flight attitude measurement unit detects that unmanned aerial vehicle flight gesture reaches and predetermines the change numerical value, the control unit record storage key position data, and image acquisition unit catches the image data of key position, establishes the flight database of unmanned aerial vehicle tracking flight.
Further, the neural network processing unit adopts a convolutional neural network to perform learning type data processing, and comprises an embedded neural Network (NPU) processor and an Intel neural network processor, wherein the NPU processor can perform deep learning and accumulate and identify image routes.
Further, the neural network processing unit comprises at least one input layer, at least two convolution layers, at least two downsampling layers, at least one full connection layer, and at least one output layer;
the processing steps of the neural network processing unit include:
inputting 160x120 resolution gray scale images into a neural network processing unit through the input layer;
after the image of the input layer passes through the first convolution layer, outputting a feature map containing at least 6 image features, wherein the size of an adopted convolution window is 5x5, and the size of the output feature map is 156x 116;
the image of the first convolution layer is downsampled through a next second sampling layer, and a feature map containing at least 6 image features is output, wherein the size of the feature map is 78x 58;
the second downsampling layer outputs at least 12 feature maps through a third convolution layer, the size of the convolution kernel is 5x5, and the size of the feature map is 74x 54;
the third convolution layer outputs at least 12 characteristic maps through a fourth down-sampling layer, and the size of the characteristic maps is 37x 27;
the fifth layer is a full connecting layer with 11988 units in total, and is respectively connected with corresponding units of the fourth downsampling layer;
the activation function used for the first layer through the fifth layer is a hyperbolic tangent function.
Further, the flight attitude measurement unit comprises at least one first substrate, at least one micro gyroscope body and at least one magnetic resistance plate body;
one side of the first substrate is sequentially stacked with the micro-gyroscope body and the magnetic resistance plate body, the micro-gyroscope body is provided with a back-bending electric coil providing a constant magnetic field, the magnetic resistance plate body is provided with a tunnel magnetic resistance piece sensitive to external magnetic field changes, the back-bending electric coil and the tunnel magnetic resistance piece are arranged oppositely, and angular rate signal detection is carried out through a tunnel magnetic resistance principle.
Further, the micro gyroscope body comprises at least one supporting frame body, at least one starting block, at least one detecting block and at least one reverse bending coil;
the supporting frame body is a frame body structure used for supporting the starting block and the detection block, the supporting frame body is located on the outermost layer of the micro gyroscope body, the starting block and the detection block are arranged in the supporting frame body, and the starting electrode, the starting feedback electrode and the inflection electric coil electrode are deposited on the supporting frame body.
Furthermore, the support frame body is connected with the starting block through a starting electrode;
the detection block is arranged at the central position of the micro gyroscope body and is connected with the starting block through the detection beam, and the return electric coil is deposited above the surface of the detection block.
Furthermore, the bent body of the continuous bent structure is formed at the position of the bent electric coil corresponding to the detection block, and the bent electric coil is connected with the support frame body through a bent electric coil electrode.
Furthermore, the starting block is arranged between the detection block and the support frame body through the starting beam;
the four corners of the starting block are respectively provided with a starting beam, and the starting beam comprises a first starting beam, a second starting beam and a starting beam connecting block;
the first starting beam and the second starting beam are long and thin beam structures, the length of the beams is far greater than the width of the beams, the first starting beam and the second starting beam are respectively positioned on two sides of the starting beam connecting block and are parallel to each other, the starting block and the starting beam connecting block are connected, the starting beam connecting block is integrally of a T-shaped structure and is connected with the first starting beam, the second starting beam and the support frame body;
the detection device comprises a detection block, a starting block, a detection beam, a first detection beam, a second detection beam and a detection beam connecting block, wherein the detection beam is arranged at the peripheral corners of the detection block and connected with the starting block;
the first detection beam and the second detection beam are in a slender beam structure, the length of the beam is far greater than the width of the beam, and the first detection beam and the second detection beam are respectively arranged on two sides of the detection beam connecting block, are parallel to each other and are connected with the detection block and the detection beam connecting block;
the whole detection beam connecting block is of a T-shaped structure and is connected with the first detection beam, the second detection beam and the starting block.
An intelligent tracking flight method of an unmanned aerial vehicle comprises the following steps:
s1 learning flight path: flying along the control route by an unmanned aerial vehicle, capturing image data along the route by the unmanned aerial vehicle, performing learning training on the image data through a neural network processing unit, and memorizing the flying route;
s2 key location learning: in the flight process of the unmanned aerial vehicle, flight attitudes of three flight dimensions of the unmanned aerial vehicle are detected through a flight attitude measuring unit, and when detection data of the flight attitude measuring unit reach preset variation, the image acquisition unit captures the surrounding flight environment and the flight attitude of the unmanned aerial vehicle;
s3 neural network processing analysis: image data captured by the unmanned aerial vehicle are all input into the neural network processing unit for deep analysis and learning, and a flight route database is built in the neural network processing unit;
s4 unmanned aerial vehicle tracking flight: the unmanned aerial vehicle carries out tracking flight according to the flight route of the flight route data, carries out image processing analysis in real time through the neural network processing unit according to the image captured by the image acquisition unit in the flight process, adjusts the self-balancing posture of the unmanned aerial vehicle and completes the whole tracking flight.
Further, in step S2, the cameras are wide-angle cameras set up at upper and lower positions, and during normal flight of the unmanned aerial vehicle, only one camera is used for capturing a flight environment image;
when the flight attitude measurement unit detects the variation of the angular rate of the unmanned aerial vehicle in real time and reaches the preset variation numerical value of the control unit, the control unit starts another camera to work, and captures the current image in a three-dimensional manner, the two cameras work simultaneously, the process image of the inertial attitude of the unmanned aerial vehicle from the variation attitude to the recovery normal flight attitude is acquired, and the position node and the time node corresponding to the neural network processing unit are input to perform recording and tracking analysis.
The invention has the advantages that: the invention has the advantages that: in order to solve the problem of autonomous tracking flight of the unmanned aerial vehicle, the tracking flight method of the unmanned aerial vehicle based on the convolutional neural network is provided, and tracking flight of the unmanned aerial vehicle is realized from a brand-new angle. Firstly, an image in front of an unmanned aerial vehicle is collected through an airborne camera, then the collected image is sent to an improved convolutional neural network model for classification, and a flight instruction is given, so that the unmanned aerial vehicle can realize a tracking function, and in the flight process, a flight attitude measurement unit based on a tunnel magnetic resistance effect with high sensitivity is adopted, and a micro gyroscope body of the flight attitude measurement unit is used for in-plane detection, so that the unmanned aerial vehicle has the advantages of small damping effect, high precision and the like compared with a micro gyroscope for out-of-plane detection;
furthermore, the zigzag electric coil is applied to the micro gyroscope, when the magnetic field sensed by the tunnel magnetic resistance element changes, the resistance value of the tunnel magnetic resistance element can change violently under the weak magnetic field change, and the detection precision of the micro gyroscope can be improved by one to two orders of magnitude. Flight attitude measuring unit realizability is strong, interface circuit simply just detects easily, can solve angular rate signal detection on being applied to unmanned aerial vehicle, through flight attitude measuring unit acquires unmanned aerial vehicle at the change key position of flight attitude, and starts image acquisition unit carries out the focus and catches the image, forms the three-dimensional model of flight environment, provides the key position image for unmanned aerial vehicle tracking flight route, guarantees that unmanned aerial vehicle is at the reasonable flight of key position, improves the whole security of unmanned aerial vehicle tracking flight.
Drawings
Fig. 1 is a schematic diagram of a working flow of a neural network processing unit according to a first embodiment of the present invention;
FIG. 2 is a schematic view of the overall structure of the flight attitude measurement unit according to the present invention;
FIG. 3 is a schematic view of the overall structure of the first substrate according to the present invention;
FIG. 4 is a schematic view of a layered structure of a flight attitude measurement unit according to the present invention;
FIG. 5 is a schematic structural diagram of a micro-gyroscope according to the invention;
FIG. 6 is a top view of a micro-gyroscope according to the invention;
FIG. 7 is a schematic structural diagram of the starting block according to the present invention;
FIG. 8 is a schematic view of a first actuating beam according to the present invention;
FIG. 9 is a schematic view of the structure of the detection block according to the present invention;
FIG. 10 is a schematic view of a first sensing beam according to the present invention;
FIG. 11 is a schematic structural diagram of a metal wire according to the present invention;
fig. 12 is a schematic data processing diagram of the control unit of the drone;
fig. 13 is a schematic diagram of a shaded portion of the unmanned aerial vehicle with the visual axis direction of the camera located on the right side of the figure;
fig. 14 is a schematic view of the view axis direction of the camera of the drone being located in the middle of two shadows in the figure;
fig. 15 is a schematic view of a hatched portion on the left side of the figure showing the direction of the visual axis of the camera of the drone;
fig. 16 is a training flowchart of the CNN convolutional neural network.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
On the contrary, the invention is intended to cover alternatives, modifications, equivalents and alternatives which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of the present invention, certain specific details are set forth in order to provide a better understanding of the present invention. It will be apparent to one skilled in the art that the present invention may be practiced without these specific details.
The invention provides an unmanned aerial vehicle intelligent tracking flight system based on a tunnel magneto-resistive micro gyroscope, wherein an image acquisition unit for image acquisition is arranged on one side of an unmanned aerial vehicle facing a flight direction, the image acquisition unit is connected with a control unit of the unmanned aerial vehicle, a flight attitude measurement unit for detecting flight inertia is arranged in the unmanned aerial vehicle, the flight attitude measurement unit is connected with the control unit, flight inertia data of the unmanned aerial vehicle are acquired in real time and are transmitted back to the control unit, and the control unit of the unmanned aerial vehicle controls each rotor wing of the unmanned aerial vehicle to perform corresponding flight actions.
The control unit adopts the convolutional neural network to perform learning type data processing, and the convolutional neural network is modified on the basis of LeNe-5 in consideration of the fact that a complex network structure not only increases training difficulty, but also has extremely high requirements on airborne hardware equipment, and the real-time performance of the system is directly influenced, and the improved LeNet-5 structure is shown in figure 1. The system comprises an input layer, two convolution layers, two down-sampling layers, a full-connection layer and an output layer.
The input data is a 160x120 resolution gray scale image, the first convolutional layer (C1) contains 6 feature maps, the size of the convolution window used is 5x5, and the size of the output feature map is 156x 116. The second downsampling layer (S2) is used to downsample the (C1) layer, and similarly obtains 6 feature maps, the size of which is 78 × 58. The third convolutional layer (C3) includes 12 feature maps, the size of the convolutional kernel is 5x5, and the size of the feature map is 74x 54; the fourth down-sampling layer (S4) includes 12, the size of the feature map is 37x 27; the fifth layer has 11988 units, which are respectively connected with the corresponding units of the fourth layer; the output layer is a fully connected layer, having three cells in total. Wherein the activation function used for the first layer through the fifth layer is a tanh function.
In addition, the classifier employed here at the final output layer of the convolutional neural network is a Softmax regression model in which, for a given test input t, the function p (y ═ j | X) is assumed to be the probability value that X belongs to the class j.
As shown in fig. 2, the flying attitude measurement unit includes a first substrate 1, a micro gyroscope 8, and a magnetic resistance plate 23, where the magnetic resistance plate 23 is provided with a tunnel magnetic resistance element 24 and an adhesion point 25, and the tunnel magnetic resistance element 24 has a plurality of pins 32 for providing an input voltage and leading out a detection signal. The micro gyroscope body 8 is manufactured by processing a silicon wafer through an MEMS (micro electro mechanical system) process and is arranged at the center above the first substrate 1, the magnetic resistance plate body 23 is arranged above the micro gyroscope body 8, and the micro gyroscope body, the magnetic resistance plate body and the magnetic resistance plate body can be integrated together through the bonding salient point 4 and the adhesion point 25.
As shown in fig. 3, the first substrate 1 is entirely square, the material can be silicon, ceramic, glass and other materials, the first substrate is provided with two starting magnets, namely a first starting magnet 2 and a second starting magnet 3, the sizes of the two starting magnets are consistent, the two starting magnets are embedded in the first substrate 1, the overall structure is a cuboid, the length and the width of the cuboid are larger than the thickness of the cuboid, and the cuboid is made of neodymium iron boron permanent magnets and used for providing a starting magnetic field for the micro gyroscope. The bonding bumps 4 are respectively arranged around the starting magnet, and metal titanium, copper and tin can be evaporated by adopting an evaporation process for bonding with the micro gyroscope body, wherein the evaporation process is a conventional process in the field and is not specifically limited herein.
As shown in fig. 4, the magnetoresistive plate 23 is provided with a tunnel magnetoresistive element 24 and an adhesion point 25, when a voltage is input to the tunnel magnetoresistive element 24 through a pin 32, the tunnel magnetoresistive element 24 is sensitive to a change in an external magnetic field, and a resistance value of the tunnel magnetoresistive element 24 changes dramatically, and a detected signal is output through an internal wheatstone bridge and the pin 32. The adhesion points 25 are disposed at the peripheral edges of the magnetoresistive plate body, and can be made of ceramic, epoxy resin, etc., for bonding with the micro gyroscope body.
As shown in fig. 5 and 6, the micro gyroscope body 8 can be divided into three parts, namely a support frame body 5, an actuating block 6, a detection block 7, and an actuating connecting lead 26, an actuating feedback lead 27 and a reverse bending electric coil 28 which are arranged on the upper surface. The micro gyroscope body is manufactured by adopting an MEMS silicon processing technology, the metal conducting wire can be manufactured by adopting a magnetron sputtering technology, an evaporation technology and an electroplating technology, and the material can be copper, silver, gold, aluminum and other metals. The support frame body 5 is positioned at the outermost layer of the micro gyroscope body and is used for supporting the starting block 6 and the detection block 7, and the starting electrode 29, the starting feedback electrode 30 and the reverse bending coil electrode 31 are deposited above the support frame body.
The starting block 6 is arranged between the detection block 7 and the support frame body 5, the first starting beam 9, the second starting beam 10, the third starting beam 11 and the fourth starting beam 12 are connected with the external support frame body 5, and the upper portion of the lower portion of the upper portion of the lower. The detection block 7 is arranged at the central position of the micro gyroscope, is connected with the starting block 6 through a first detection beam 16, a second detection beam 17, a third detection beam 18 and a fourth detection beam 19, and is deposited with a back-folded electric coil 28.
The starting connecting wire 26 is arranged on the upper surface of the left side of the starting block 6 and is connected with a starting electrode 29 on the support frame body 5 through the first starting beam 13 and the starting connecting block 15; the starting feedback lead 27 is arranged on the upper surface of the right side of the starting block 6, has the same size as the starting connecting lead 26, and is connected with a starting feedback electrode 30 on the support frame body 5 through the second starting beam 24 and the starting connecting block 15; the inflection electric coil 28 is arranged on the upper surface of the detection block 7 and is connected with an inflection electric coil electrode 31 on the support frame body 5 through a second detection beam 21, a detection connection block 22, a starting block 6, a second starting beam 14 and a starting connection block 15.
As shown in fig. 7 and 8, four corners of the starting block 6 are provided with a first starting beam 9, a second starting beam 10, a third starting beam 11 and a fourth starting beam 12, the first starting beam 9, the second starting beam 10, the third starting beam 11 and the fourth starting beam 12 have the same structural size, and are composed of a first starting beam 13, a second starting beam 14 and a starting beam connecting block 15, the first starting beam 13 and the second starting beam 14 have the same structural size, and are in long and thin beam structures, i.e., the length of the beams is far greater than the width of the beams, and the first starting beam 13 and the second starting beam 14 are respectively located on two sides of the starting beam connecting block 15 and are parallel to each other for connecting the starting block 6 and the starting beam connecting block 15, and the starting beam connecting block 15 is integrally in a "T" shape for connecting the first starting beam 13, the second starting beam 14 and the support body 5.
As shown in fig. 9 and 10, the four corners of the detecting block 7 are provided with a first detecting beam 16, a second detecting beam 17, a third detecting beam 18 and a fourth detecting beam 19 which are connected with the starting block 6, the first detection beam 16, the second detection beam 17, the third detection beam 18 and the fourth detection beam 19 have the same structural size, and are composed of a first detection beam 20, a second detection beam 21 and a detection beam connecting block 22, the first detection beam 20 and the second detection beam 21 have the same structural size, and are integrally in a slender beam structure, namely, the length of the beam is much longer than the width of the beam, and the first detection beam 20 and the second detection beam 21 are respectively arranged at two sides of the detection beam connecting block 22 and are parallel to each other, the detection beam connecting block 22 is used for connecting the detection block 7 and the detection beam connecting block 22, and the detection beam connecting block 22 is integrally T-shaped and is used for connecting the first detection beam 20, the second detection beam 21 and the starting block 6.
As shown in fig. 11, the metal wires mainly include a start connecting wire 26, a start feedback wire 27, and a reverse bending coil 28. The upper side and the lower side of the starting connecting wire 26 are connected with a starting electrode 29, and when current is introduced to the starting electrode 29, the starting connecting wire 26 generates ampere force under the action of a magnetic field to provide starting force for the micro gyroscope. Start feedback electrode 30 is connected to both sides about starting feedback wire 27, and when micro-gyroscope along the direction of starting (X axle) reciprocating motion, start feedback wire 27 and can cut the magnetic field line, and self can produce electric current, starts to realize amplitude stabilization control to micro-gyroscope through the electric current size that detects start feedback electrode 30. When a certain current is applied to the electrode 31 of the folded electric coil, the folded electric coil 28 can generate a constant magnetic field in the starting direction of the micro gyroscope, can generate a sinusoidally-varying magnetic field in the detection direction of the micro gyroscope, and can control the magnitude of the magnetic field by changing the magnitude of the current input into the folded electric coil 28.
The micro gyroscope body 8 of the invention is arranged above a first substrate 1, in a uniform magnetic field generated by a first starting magnet 2 and a second starting magnet 3, an alternating starting current is loaded on a starting connecting wire to generate an alternating Lorentz force, a starting block 6 vibrates in a reciprocating manner along a starting direction (X axis) under the action of the starting force, when an unmanned aerial vehicle is accelerated uniformly or rotates with large angular velocity of corner inertia, a detection block 7 moves along a detection direction (Y axis) under the action of the Coriolis force when the angular velocity of the Z axis direction is input, the detection block 7 drives a folding electric coil 28 to oscillate in a stable manner relative to a tunnel magnetic resistance piece 24 fixed on a magnetic resistance plate body 23, when the magnetic field sensed by the tunnel magnetic resistance piece 24 changes, the resistance value of the detection block greatly changes, the detection of the angular velocity can be realized by measuring the resistance value change, the flight attitude measuring unit transmits accurately collected angular velocity data to a control unit in real time,
as shown in fig. 12, the data processing of the control unit of the drone: in order to train the convolutional neural network, images around the flight path of the unmanned aerial vehicle need to be acquired, and the images are preprocessed and calibrated. Is provided with
The correct direction in which the drone should travel, flying in this direction may keep the drone on the correct path,
for the current flight direction of the drone, i.e. the pointing direction of the camera, and consider
Parallel to the horizontal plane. Let a be a vector
Sum vector
The included angle therebetween.
As shown in fig. 13, when 15 ° < a <90 °, i.e. the viewing axis direction of the camera is located in the right shaded portion of the figure, the drone should turn left to remain on the correct path, with the image captured by the camera being marked TL.
As shown in fig. 14, when-15 ° < a <15 °, i.e. the camera's boresight direction is in the middle of the two shades in the figure, the drone can be advanced, with the image captured by the camera marked GS.
As shown in fig. 15, when a 90 ° < a < -15 °, i.e. the camera's boresight direction is located in the left shaded portion of the figure, the drone should be turned right to remain in the correct path, the image captured by the camera being now marked TR.
In order to obtain the training data set, the unmanned aerial vehicle is controlled by an operator to fly in a corridor through a remote controller, and during the flight, images are captured every 0.2 meter of forward flight, and 500 images marked as TL, 500 images marked as GS and 500 images marked as TR are obtained in total. In addition, in order to expand the training data set, the training set is expanded to 3000 images by mirror-inverting the images. After the mirror image is flipped, the image originally labeled GS is now also labeled GS, and the images originally labeled TL, TR are now labeled TR, TL.
Further, the image acquisition unit comprises at least two cameras for capturing images, the cameras are wide-angle cameras arranged at the upper and lower positions, in the normal flight of the unmanned aerial vehicle, only one camera is used for image capture, the flight attitude of the unmanned aerial vehicle in three-axis dimensions is detected in real time by the flight attitude measurement unit, when the three-axis flight attitude of the unmanned aerial vehicle has larger data change and the variation of the angular rate reaches the preset variation value of the control unit, the control unit records the attitude of the unmanned aerial vehicle as a key position at the moment, the other camera starts to work, and the current image is captured stereoscopically, the two cameras work simultaneously to obtain the process image of the inertial attitude of the unmanned aerial vehicle from the changing attitude to the recovering normal flight attitude, inputting the position node and the time node corresponding to the neural network processing unit to perform recording and tracking analysis;
for example, when the unmanned aerial vehicle encounters a turning flight scene, a rapid lifting flight scene and other flight scenes in the flight process, two cameras are started simultaneously in the flight training learning process, a three-dimensional model of the current flight environment of the unmanned aerial vehicle can be acquired in real time through the two wide-angle cameras, the flight attitude data of the position of the key and the recovery action of the attitude adjustment are recorded, and a flight route database is provided for the return flight or the retracing flight of the unmanned aerial vehicle.
As shown in fig. 16, which is a training flow chart of CNN convolutional neural network, it is essentially an input-to-output mapping, and it can learn the mapping relationship between a large number of input persons and output without any precise mathematical expression between input persons and output, and the network has the mapping capability between pairs of input persons and output by training the convolutional network with a known pattern.
Further, the convolutional neural network is relatively complex, and a training process requires a large amount of computation, after the camera of the unmanned aerial vehicle acquires image data, the camera may analyze data of the image through the neural network processing unit, where the neural network processing unit is specifically a neural network processing chip, and the neural network processing chip includes, but is not limited to, an embedded neural Network (NPU) processor, an intel neural network processor, and the like, which may perform deep learning and accumulate and identify an image route, and is not limited specifically herein.
In this embodiment, the image acquisition unit is connected with the neural processing unit, the neural network processing unit is connected with the control unit, the control unit controls the motors of the rotors of the unmanned aerial vehicle and controls the overall flight attitude of the unmanned aerial vehicle, the control unit includes but is not limited to a single chip microcomputer, a microprocessor or a PCB with a control circuit, and only a conventional processor for realizing data storage and analysis functions is required, and no specific limitation is imposed herein.
Unmanned aerial vehicle flight test and result analysis: the activation function used by the three neurons of the last output layer of the neural network is Softmax, the output value of which can be considered as the probability that the current input image belongs to each class (TL, TR, GS), and therefore it is necessary to convert the output value into a control signal for controlling the flight of the drone, wherein the heading angle of the drone is proportional to/K71) -PCTR, the value of which is positive the drone turns to the left, and the value of which is negative the drone turns to the right, the distance D that the drone flies forward is proportional to p (GS), and D/p (GS) is 0.2, that is, the drone flies 0.2 meters forward when p (GS) is 1.
Further, in a second embodiment of the present invention, the second embodiment includes a method for intelligent tracked flight of a drone, the method including the steps of:
s1 learning flight path: flying along the control route by an unmanned aerial vehicle, capturing image data along the route by the unmanned aerial vehicle, performing learning training on the image data through a neural network processing unit, and memorizing the flying route;
s2 key location learning: in the flight process of the unmanned aerial vehicle, flight attitudes of three flight dimensions of the unmanned aerial vehicle are detected through a flight attitude measuring unit, and when detection data of the flight attitude measuring unit reach preset variation, the image acquisition unit captures the surrounding flight environment and the flight attitude of the unmanned aerial vehicle;
s3 neural network processing analysis: image data captured by the unmanned aerial vehicle are all input into the neural network processing unit for deep analysis and learning, and a flight route database is built in the neural network processing unit;
s4 unmanned aerial vehicle tracking flight: the unmanned aerial vehicle carries out tracking flight according to the flight route of the flight route data, and carries out image processing and analysis in real time through the neural network processing unit according to the image captured by the image acquisition unit in the flight process, so that the self-balancing posture of the unmanned aerial vehicle is adjusted, and the whole tracking flight is completed.
Further, in step S2, the camera is a wide-angle camera disposed at an up-down position, and during normal flight of the unmanned aerial vehicle, only one camera is used for capturing a flight environment graph;
when the flight attitude measurement unit detects the variation of the angular rate of the unmanned aerial vehicle in real time and reaches the preset variation numerical value of the control unit, the control unit starts another camera to work, and captures the current image in a three-dimensional manner, the two cameras work simultaneously, the process image of the inertial attitude of the unmanned aerial vehicle from the variation attitude to the recovery normal flight attitude is acquired, and the position node and the time node corresponding to the neural network processing unit are input to perform recording and tracking analysis.
The invention has the advantages that: in order to solve the problem of autonomous tracking flight of the unmanned aerial vehicle, the tracking flight method of the unmanned aerial vehicle based on the convolutional neural network is provided, and tracking flight of the unmanned aerial vehicle is realized from a brand new angle. Firstly, an image in front of an unmanned aerial vehicle is collected through an airborne camera, then the collected image is sent to an improved convolutional neural network model for classification, and a flight instruction is given, so that the unmanned aerial vehicle can realize a tracking function, and in the flight process, a flight attitude measurement unit based on a tunnel magnetic resistance effect with high sensitivity is adopted, and a micro gyroscope body of the flight attitude measurement unit is used for in-plane detection, so that the unmanned aerial vehicle has the advantages of small damping effect, high precision and the like compared with a micro gyroscope for out-of-plane detection;
furthermore, the zigzag electric coil is applied to the micro gyroscope, when the magnetic field sensed by the tunnel magnetic resistance element changes, the resistance value of the tunnel magnetic resistance element can change violently under the weak magnetic field change, and the detection precision of the micro gyroscope can be improved by one to two orders of magnitude. Flight attitude measuring unit realizability is strong, interface circuit simply just detects easily, can solve angular rate signal detection on being applied to unmanned aerial vehicle, through flight attitude measuring unit acquires unmanned aerial vehicle at the change key position of flight attitude, and starts image acquisition unit carries out the focus and catches the image, forms the three-dimensional model of flight environment, provides the key position image for unmanned aerial vehicle tracking flight route, guarantees that unmanned aerial vehicle is at the reasonable flight of key position, improves the whole security of unmanned aerial vehicle tracking flight.
It will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in the embodiments described above without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims.
Claims (9)
1. An intelligent tracking flight system of an unmanned aerial vehicle is applied to the flight of the unmanned aerial vehicle and is characterized in that the flight system comprises an image acquisition unit for acquiring images, a flight attitude measurement unit for detecting flight inertia, a neural network processing unit for processing a neural network and a control unit for controlling the flight action of the unmanned aerial vehicle;
the image acquisition unit is connected with the neural network processing unit and transmits acquired image data to the neural network processing unit, the flight attitude measurement unit detects the flight attitude of the unmanned aerial vehicle in real time, the control unit records and stores the position data of the key positions when the flight attitude measurement unit detects that the flight attitude of the unmanned aerial vehicle reaches a preset change value, the image acquisition unit captures the image data of the key positions and establishes a flight database for the tracking flight of the unmanned aerial vehicle,
the flight attitude measurement unit comprises at least one first substrate, at least one micro gyroscope body and at least one magnetic resistance plate body;
one side of first base plate stacks gradually the setting the little top body, and the magnetic resistance plate body, the little top body has the inflection electric coil that provides invariable magnetic field, the magnetic resistance plate body has the tunnel magnetic resistance piece that sensitive external magnetic field changes, inflection electric coil and tunnel magnetic resistance piece set up relatively to carry out angular rate signal detection through tunnel magnetic resistance principle.
2. The intelligent tracking flight system for the unmanned aerial vehicle as claimed in claim 1, wherein the neural network processing unit performs learning type data processing by using a convolutional neural network, and the neural network processing unit comprises an embedded neural network NPU processor and an Intel neural network processor which can perform deep learning and accumulate and identify image routes.
3. The unmanned aerial vehicle intelligent tracking flight system of claim 1, wherein the neural network processing unit comprises at least one input layer, at least two convolutional layers, at least two downsampling layers, at least one fully-connected layer, at least one output layer;
the processing steps of the neural network processing unit include:
inputting 160x120 resolution gray scale image into the neural network processing unit through the input layer;
after the image of the input layer passes through the first convolution layer, outputting a feature map containing at least 6 image features, wherein the size of an adopted convolution window is 5x5, and the size of the output feature map is 156x 116;
the image of the first convolution layer is downsampled through a next second sampling layer, and a feature map containing at least 6 image features is output, wherein the size of the feature map is 78x 58;
the second downsampling layer passes through a third convolutional layer, the output of the second downsampling layer comprises at least 12 feature maps, the size of a convolutional kernel is 5x5, and the size of a feature map is 74x 54;
the third convolution layer outputs at least 12 feature maps through a fourth down-sampling layer, wherein the size of the feature maps is 37x 27;
the fifth layer is a full-connection layer with 11988 units in total and is connected with the corresponding unit of the fourth downsampling layer;
the activation function used for the first layer through the fifth layer is a hyperbolic tangent function.
4. The intelligent tracking flight system of the unmanned aerial vehicle as claimed in claim 1, wherein the micro gyroscope body comprises at least one support frame body, at least one starting block, at least one detecting block, and at least one bending coil;
the supporting frame body is a frame body structure used for supporting the starting block and the detection block, the supporting frame body is located on the outermost side of the micro gyroscope body, the starting block and the detection block are arranged in the supporting frame body, and the starting electrode, the starting feedback electrode and the inflection electric coil electrode are deposited on the supporting frame body.
5. The intelligent unmanned aerial vehicle tracking flight system of claim 4, wherein the support frame body is connected with the starting block through a starting electrode;
the detection block is arranged at the central position of the micro gyroscope body and is connected with the starting block through the detection beam, and the return electric coil is deposited above the surface of the detection block.
6. The intelligent tracking flight system of the unmanned aerial vehicle as claimed in claim 4, wherein the bent body of the continuous bent structure is formed at a position of the detection block corresponding to the bent electric coil, and the bent electric coil is connected with the support frame body through a bent electric coil electrode.
7. The intelligent tracking flight system for unmanned aerial vehicles according to claim 4, wherein the starting block is arranged between the detection block and the support frame body through a starting beam;
the four corners of the starting block are respectively provided with a starting beam, and the starting beam comprises a first starting beam, a second starting beam and a starting beam connecting block;
the first starting beam and the second starting beam are long and thin beam structures, the length of the beams is far larger than the width of the beams, the first starting beam and the second starting beam are respectively positioned on two sides of the starting beam connecting block and are parallel to each other, the starting block and the starting beam connecting block are connected, the starting beam connecting block is integrally of a T-shaped structure and is connected with the first starting beam, the second starting beam and the support frame body;
the detection device comprises a detection block, a starting block, a detection beam, a first detection beam, a second detection beam and a detection beam connecting block, wherein the detection beam is arranged at the peripheral corners of the detection block and connected with the starting block;
the first detection beam and the second detection beam are in a slender beam structure, the length of the beam is far greater than the width of the beam, and the first detection beam and the second detection beam are respectively arranged on two sides of the detection beam connecting block, are parallel to each other and are connected with the detection block and the detection beam connecting block;
the whole detection beam connecting block is of a T-shaped structure and is connected with the first detection beam, the second detection beam and the starting block.
8. An intelligent tracking flight method of an unmanned aerial vehicle, which is applied to the intelligent tracking flight system of the unmanned aerial vehicle according to any one of claims 1 to 7, and is characterized by comprising the following steps:
s1 learning flight path: flying along the control route by an unmanned aerial vehicle, capturing image data along the route by the unmanned aerial vehicle, performing learning training on the image data through a neural network processing unit, and memorizing the flying route;
s2 key location learning: in the flight process of the unmanned aerial vehicle, flight attitudes of three flight dimensions of the unmanned aerial vehicle are detected through a flight attitude measuring unit, and when detection data of the flight attitude measuring unit reach preset variation, the image acquisition unit captures the surrounding flight environment and the flight attitude of the unmanned aerial vehicle;
s3 neural network processing analysis: image data captured by the unmanned aerial vehicle are all input into the neural network processing unit for deep analysis and learning, and a flight route database is built in the neural network processing unit;
s4 unmanned aerial vehicle tracking flight: the unmanned aerial vehicle carries out tracking flight according to the flight route of the flight route data, carries out image processing analysis in real time through the neural network processing unit according to the image captured by the image acquisition unit in the flight process, adjusts the self-balancing posture of the unmanned aerial vehicle and finishes the whole tracking flight.
9. The intelligent tracking flight method for the unmanned aerial vehicle as claimed in claim 8, wherein in step S2, the image acquisition unit is a camera, the camera is a wide-angle camera disposed at an upper position and a lower position, and during normal flight of the unmanned aerial vehicle, only one camera is used for capturing a flight environment image;
when the flight attitude measurement unit detects the variation of the angular rate of the unmanned aerial vehicle in real time and reaches the preset variation numerical value of the control unit, the control unit starts another camera to work, and captures the current image in a three-dimensional manner, the two cameras work simultaneously, the process image of the inertial attitude of the unmanned aerial vehicle from the variation attitude to the recovery normal flight attitude is acquired, and the position node and the time node corresponding to the neural network processing unit are input to perform recording and tracking analysis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548062.9A CN110456805B (en) | 2019-06-24 | 2019-06-24 | Intelligent tracking flight system and method for unmanned aerial vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910548062.9A CN110456805B (en) | 2019-06-24 | 2019-06-24 | Intelligent tracking flight system and method for unmanned aerial vehicle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110456805A CN110456805A (en) | 2019-11-15 |
| CN110456805B true CN110456805B (en) | 2022-07-19 |
Family
ID=68480804
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910548062.9A Active CN110456805B (en) | 2019-06-24 | 2019-06-24 | Intelligent tracking flight system and method for unmanned aerial vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110456805B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222726A (en) * | 2019-05-15 | 2019-09-10 | 北京字节跳动网络技术有限公司 | Image processing method, device and electronic equipment |
| CN110991502B (en) * | 2019-11-21 | 2020-10-13 | 北京航空航天大学 | Airspace security situation assessment method based on category activation mapping technology |
| CN110908399B (en) * | 2019-12-02 | 2023-05-12 | 广东工业大学 | Unmanned aerial vehicle autonomous obstacle avoidance method and system based on lightweight neural network |
| CN111142557B (en) * | 2019-12-23 | 2020-12-22 | 清华大学 | Unmanned aerial vehicle path planning method and system, computer equipment and readable storage medium |
| CN110879059B (en) * | 2019-12-30 | 2022-05-27 | 中北大学 | Tunnel magnetoresistance effect micro-gyroscope device and method based on piezoelectric ceramic out-of-plane driving |
| CN111572787B (en) * | 2020-05-07 | 2022-05-03 | 南通大学 | Unmanned aerial vehicle self-balancing device based on planetary gear speed change and balancing method thereof |
| CN113284144B (en) * | 2021-07-22 | 2021-11-30 | 深圳大学 | Tunnel detection method and device based on unmanned aerial vehicle |
| CN114063496A (en) * | 2021-11-02 | 2022-02-18 | 广州昂宝电子有限公司 | Unmanned aerial vehicle control method and system and remote controller for remotely controlling unmanned aerial vehicle |
| CN114354082B (en) * | 2022-03-18 | 2022-05-31 | 山东科技大学 | Intelligent tracking system and method for submarine pipeline based on imitated sturgeon whisker |
| CN115826628B (en) * | 2023-02-22 | 2023-05-09 | 成都航空职业技术学院 | Heterogeneous unmanned aerial vehicle vision obstacle avoidance system and method based on NeRF neural network |
| CN116630828B (en) * | 2023-05-30 | 2023-11-24 | 中国公路工程咨询集团有限公司 | Unmanned aerial vehicle remote sensing information acquisition system and method based on terrain environment adaptation |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102636995A (en) * | 2012-05-03 | 2012-08-15 | 河海大学常州校区 | Method for controlling micro gyro based on radial basis function (RBF) neural network sliding mode |
| CN104281155A (en) * | 2014-11-07 | 2015-01-14 | 中国人民解放军国防科学技术大学 | Three-dimensional flight path tracking method for unmanned airship |
| CN205229800U (en) * | 2015-11-21 | 2016-05-11 | 浙江水利水电学院 | Intelligence tracking car based on BP neural network PID controlling means |
| CN106483969A (en) * | 2016-12-16 | 2017-03-08 | 北京中科浩电科技有限公司 | A kind of many rotors vision tracking unmanned plane |
| CN106873609A (en) * | 2015-12-10 | 2017-06-20 | 北京航天长峰科技工业集团有限公司 | The identification of unmanned plane aerial statue and control method based on quaternary number and neutral net |
| CN107677287A (en) * | 2017-09-22 | 2018-02-09 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and dolly based on convolutional neural networks follow line method |
| CN107703936A (en) * | 2017-09-22 | 2018-02-16 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and dolly localization method based on convolutional neural networks |
| WO2018128671A1 (en) * | 2017-01-03 | 2018-07-12 | Qualcomm Incorporated | Capturing images of a game by an unmanned autonomous vehicle |
| CN108603348A (en) * | 2016-02-10 | 2018-09-28 | 爱知制钢株式会社 | The construction method and working truck system of magnetic mark |
| WO2019024303A1 (en) * | 2017-08-02 | 2019-02-07 | 华南理工大学 | Stable flight control method for multi-rotor unmanned aerial vehicle based on finite-time neurodynamics |
| CN109407698A (en) * | 2018-10-25 | 2019-03-01 | 深圳慈航无人智能系统技术有限公司 | No-manned plane three-dimensional path barrier-avoiding method based on probability density |
| WO2019055690A1 (en) * | 2017-09-13 | 2019-03-21 | Flirtey Holdings, Inc. | Unmanned aerial vehicle and payload delivery system |
| CN109606730A (en) * | 2018-10-25 | 2019-04-12 | 深圳慈航无人智能系统技术有限公司 | A kind of unmanned plane fault finding system and recognition methods based on quantum nerve network |
-
2019
- 2019-06-24 CN CN201910548062.9A patent/CN110456805B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102636995A (en) * | 2012-05-03 | 2012-08-15 | 河海大学常州校区 | Method for controlling micro gyro based on radial basis function (RBF) neural network sliding mode |
| CN104281155A (en) * | 2014-11-07 | 2015-01-14 | 中国人民解放军国防科学技术大学 | Three-dimensional flight path tracking method for unmanned airship |
| CN205229800U (en) * | 2015-11-21 | 2016-05-11 | 浙江水利水电学院 | Intelligence tracking car based on BP neural network PID controlling means |
| CN106873609A (en) * | 2015-12-10 | 2017-06-20 | 北京航天长峰科技工业集团有限公司 | The identification of unmanned plane aerial statue and control method based on quaternary number and neutral net |
| CN108603348A (en) * | 2016-02-10 | 2018-09-28 | 爱知制钢株式会社 | The construction method and working truck system of magnetic mark |
| CN106483969A (en) * | 2016-12-16 | 2017-03-08 | 北京中科浩电科技有限公司 | A kind of many rotors vision tracking unmanned plane |
| WO2018128671A1 (en) * | 2017-01-03 | 2018-07-12 | Qualcomm Incorporated | Capturing images of a game by an unmanned autonomous vehicle |
| WO2019024303A1 (en) * | 2017-08-02 | 2019-02-07 | 华南理工大学 | Stable flight control method for multi-rotor unmanned aerial vehicle based on finite-time neurodynamics |
| WO2019055690A1 (en) * | 2017-09-13 | 2019-03-21 | Flirtey Holdings, Inc. | Unmanned aerial vehicle and payload delivery system |
| CN107703936A (en) * | 2017-09-22 | 2018-02-16 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and dolly localization method based on convolutional neural networks |
| CN107677287A (en) * | 2017-09-22 | 2018-02-09 | 南京轻力舟智能科技有限公司 | Automatic Guided Vehicle system and dolly based on convolutional neural networks follow line method |
| CN109407698A (en) * | 2018-10-25 | 2019-03-01 | 深圳慈航无人智能系统技术有限公司 | No-manned plane three-dimensional path barrier-avoiding method based on probability density |
| CN109606730A (en) * | 2018-10-25 | 2019-04-12 | 深圳慈航无人智能系统技术有限公司 | A kind of unmanned plane fault finding system and recognition methods based on quantum nerve network |
Non-Patent Citations (5)
| Title |
|---|
| He Qing-hua ; Luo Wei ; Zou Xiang-fu.Research on the design of swing pilot platform for aerial photographic UAV.《2010 The 2nd International Conference on Industrial Mechatronics and Automation》.2010, * |
| 基于32 位单片机的四旋翼飞行器寻迹系统;曲云鹏等;《科技论坛》;20171231;第154-155页 * |
| 基于卷积神经网络的无人机循迹方法;陈思锐等;《长春理工大学学报（自然科学版）》;20180228;第110-113页 * |
| 常压封装下隧道磁阻微陀螺的高灵敏结构设计;褚伟航等;《传感技术学报》;20160331;第332-337页 * |
| 无人地效飞行器控制系统研究;李光春等;《导航与控制》;20110831;第1-6页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110456805A (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110456805B (en) | Intelligent tracking flight system and method for unmanned aerial vehicle | |
| CN105955308B (en) | The control method and device of a kind of aircraft | |
| US7440585B2 (en) | Autonomous vehicle and motion control therefor | |
| Weiss | Vision based navigation for micro helicopters | |
| EP1898291B1 (en) | Apparatus and method for estimating location of mobile body and generating map of mobile body environment using upper image of mobile body environment, and computer readable recording medium storing computer program controlling the apparatus | |
| CN106339006B (en) | A kind of method for tracking target and device of aircraft | |
| CN106197429A (en) | A kind of Multi-information acquisition location equipment and system | |
| CN106525074A (en) | Compensation method and device for holder drift, holder and unmanned aerial vehicle | |
| CN106197452A (en) | A kind of visual pattern processing equipment and system | |
| Zhang et al. | Vision-aided localization for ground robots | |
| Deng et al. | Global optical flow-based estimation of velocity for multicopters using monocular vision in GPS-denied environments | |
| Deng et al. | Visual–inertial estimation of velocity for multicopters based on vision motion constraint | |
| CN113465596A (en) | Four-rotor unmanned aerial vehicle positioning method based on multi-sensor fusion | |
| CN113256728B (en) | IMU equipment parameter calibration method and device, storage medium and electronic device | |
| CN112652001A (en) | Underwater robot multi-sensor fusion positioning system based on extended Kalman filtering | |
| CN116659490A (en) | Low cost vision-inertial fusion SLAM method | |
| CN110189359B (en) | Rapid intelligent tracking method and device based on classification learning | |
| Corke et al. | Sensing and control on the sphere | |
| Lim et al. | MSDPN: Monocular depth prediction with partial laser observation using multi-stage neural networks | |
| CN109542094A (en) | Mobile robot visual point stabilization without desired image | |
| Mirisola et al. | Trajectory recovery and 3d mapping from rotation-compensated imagery for an airship | |
| Li et al. | Motion Tracking and Detection System Based on Motion Sensor | |
| Ogata et al. | A robust position and posture measurement system using visual markers and an inertia measurement unit | |
| CN115752442B (en) | Monocular vision-based auxiliary inertial positioning method | |
| CN117437290B (en) | Multi-sensor fusion type three-dimensional space positioning method for unmanned aerial vehicle in natural protection area |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |