CN109634309B - Autonomous obstacle avoidance system and method for aircraft and aircraft - Google Patents
Autonomous obstacle avoidance system and method for aircraft and aircraft Download PDFInfo
- Publication number
- CN109634309B CN109634309B CN201910132801.6A CN201910132801A CN109634309B CN 109634309 B CN109634309 B CN 109634309B CN 201910132801 A CN201910132801 A CN 201910132801A CN 109634309 B CN109634309 B CN 109634309B
- Authority
- CN
- China
- Prior art keywords
- module
- aircraft
- flight
- aircrafts
- obstacle
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000004458 analytical method Methods 0.000 claims abstract description 38
- 238000001514 detection method Methods 0.000 claims abstract description 38
- 238000013461 design Methods 0.000 claims abstract description 25
- 238000004364 calculation method Methods 0.000 claims abstract description 7
- 239000003795 chemical substances by application Substances 0.000 claims description 39
- 238000005070 sampling Methods 0.000 claims description 31
- 238000001914 filtration Methods 0.000 claims description 25
- 230000008859 change Effects 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 16
- 239000013598 vector Substances 0.000 claims description 16
- 230000008878 coupling Effects 0.000 claims description 15
- 238000010168 coupling process Methods 0.000 claims description 15
- 238000005859 coupling reaction Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 15
- 239000002245 particle Substances 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 238000003066 decision tree Methods 0.000 claims description 9
- 230000003287 optical effect Effects 0.000 claims description 7
- 238000009928 pasteurization Methods 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000007613 environmental effect Effects 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 4
- 108010076504 Protein Sorting Signals Proteins 0.000 claims description 3
- 238000013016 damping Methods 0.000 claims description 3
- 238000007405 data analysis Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims description 2
- 238000013138 pruning Methods 0.000 claims description 2
- 230000008030 elimination Effects 0.000 claims 1
- 238000003379 elimination reaction Methods 0.000 claims 1
- 238000011160 research Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 22
- 238000003062 neural network model Methods 0.000 description 6
- 210000002569 neuron Anatomy 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000005284 excitation Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000005260 corrosion Methods 0.000 description 3
- 230000007797 corrosion Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003213 activating effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000013112 stability test Methods 0.000 description 2
- 235000015842 Hesperis Nutrition 0.000 description 1
- 235000012633 Iberis amara Nutrition 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G05D1/10—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 belongs to the technical field of aircrafts, and discloses an autonomous obstacle avoidance system and method of an aircraft and the aircraft, wherein the autonomous obstacle avoidance system comprises: the system comprises a solar power supply module, an image acquisition module, an obstacle information detection module, a central control module, an analysis and judgment module, an instruction generation module, a path planning module, a stability analysis module and a display module. When judging that other aircrafts collide with the aircraft, the path planning module adjusts the flight planning track, so that the safety flight track can be planned for the aircraft, and the flight safety is greatly improved; meanwhile, the stability analysis module can greatly reduce the calculation cost while guaranteeing the prediction precision, so that qualitative and quantitative research on the dynamic stability characteristics of the whole flight domain of the aircraft is easy to develop, and the design direction with guiding significance on the aircraft design is obtained.
Description
Technical Field
The invention belongs to the technical field of aircrafts, and particularly relates to an autonomous obstacle avoidance system and method for an aircraft and the aircraft.
Background
An aircraft (flight vehicle) is an instrument that flies within the atmosphere or outside the atmosphere (space). Aircraft fall into 3 categories: aircraft, spacecraft, rockets, and missiles. Flying in the atmosphere is known as an aircraft, such as a balloon, airship, airplane, etc. They fly by aerodynamic lift generated by static buoyancy of air or relative motion of air. In space flight, the aircraft is called a spacecraft, such as an artificial earth satellite, a manned spacecraft, a space probe, a space plane and the like. They get the necessary speed into space under the propulsion of the carrier rocket and then rely on inertia to do orbital motion similar to celestial bodies. However, existing autonomous obstacle avoidance systems for aircraft are low in safety; meanwhile, the stability test for the flight of the aircraft is complex and tedious, and the analysis error is large.
In summary, the problems of the prior art are:
(1) The existing autonomous obstacle avoidance system of the aircraft is low in safety; meanwhile, the stability test for the flight of the aircraft is complex and tedious, and the analysis error is large.
(2) In the prior art, in the process of collecting the environment image in front of the flight of the aircraft, under the condition of shielding and overlapping of multiple moving targets, the traditional target tracking algorithm is adopted, so that the tracking success rate of the moving targets cannot be effectively improved.
(3) In the prior art, in the process of detecting multiple targets on an obstacle in a front detectable distance by a millimeter wave radar and obtaining position, speed and azimuth information of the obstacle, in a complex detection background, the echo signal is difficult to avoid being doped with a plurality of clutter components, and the clutter components cannot be effectively eliminated and left in a false target mode by adopting a traditional algorithm.
(4) In the prior art, in the process of analyzing and judging the risk of the obstacle according to the acquired image and obstacle information, a decision tree of the risk degree of the obstacle is established by adopting a traditional algorithm according to the acquired image and obstacle data information, and the risk of the obstacle cannot be rapidly determined.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides an autonomous obstacle avoidance system and method for an aircraft and the aircraft.
The invention is realized in such a way that an aircraft autonomous obstacle avoidance method comprises the following steps:
step one, converting solar energy into electric energy by using a solar panel to supply power for an aircraft; an image acquisition module is used for acquiring an environment image in front of the flying of the aircraft by using an optical camera;
step two, performing multi-target detection on the obstacle in the front detectable distance by using a millimeter wave radar through an obstacle information detection module to obtain the position, speed and azimuth information of the obstacle;
step three, analyzing and judging the risk of the obstacle according to the acquired image and the obstacle information by utilizing a data analysis program;
generating a route change instruction according to the judgment result by using an instruction generator; planning a flight path of the aircraft by using a navigation system according to the change instruction through a path planning module;
fifthly, analyzing the stability of the aircraft by utilizing an analysis program;
and step six, displaying and collecting environmental images, obstacle information, flight paths and stability data by using a display.
Further, the path planning module planning method includes:
(1) Receiving a route change instruction, and generating a flight planning track for the aircraft, wherein the flight planning track comprises a flight planning route and a flight planning height layer;
(2) When other aircrafts except the aircrafts fly in the flight planning altitude layer, judging whether the other aircrafts collide with the aircrafts according to a preset rule;
(3) And when judging that the other aircrafts collide with the aircrafts, adjusting the flight planning track until judging that the other aircrafts do not collide with the aircrafts according to the preset rule.
Further, the determining whether the other aircraft collides with the aircraft according to the preset rule includes:
acquiring flight trajectories and flight speeds of the other aircrafts;
when the flight planning track and the flight track of the other aircrafts have a superposition area, calculating the time point interval of the other aircrafts and the aircrafts reaching the superposition area according to the flight track and the flight speed of the other aircrafts and the flight planning track and the flight speed of the aircrafts;
and when the time point interval is smaller than or equal to a preset time interval, judging that the other aircrafts collide with the aircrafts.
Further, the adjusting the flight planning trajectory of the aircraft includes:
adjusting the flight plan trajectory by adjusting a flight plan altitude layer and/or a flight plan route of the aircraft at least at the overlap region; the adjusted flight planning altitude of the aircraft is located between a lower limit altitude and an upper limit altitude of the aircraft.
Further, the stability analysis method includes:
1) Setting flight parameters of an aircraft;
2) Aiming at aerodynamic characteristics of an aircraft, determining target precision requirements and an upper limit of calculated amount, and selecting a predetermined number of initial sample points in a full flight domain range;
3) At an initial sample point x (i), a time sequence y (x (i), k) of unsteady aerodynamic force and moment coefficients of forced movement of the aircraft is obtained by using CFD calculation for a given set of movement training signal sequences u (x (i), k), and then a Kriging agent model is built as an initial target agent model by taking u (x (i), k) and y (x (i), k) as sample data;
4) Constructing a reference agent model, and evaluating an initial sample point to determine the current sampling precision;
5) Generating new candidate sampling points by using a test design method; calculating and evaluating all candidate sampling points through the target agent model and the reference agent model respectively to obtain precision change values generated after the candidate sampling points are added into the target agent model and the reference agent model;
6) Selecting candidate sampling points meeting the current sampling precision requirement in the precision change value as sample points added into the target agent model and the reference agent model for training through a self-adaptive sampling criterion;
7) Repeating the steps 4) to 5), judging whether the precision of the current target agent model meets the standard and whether the calculated amount reaches the upper limit or not when the current target agent model is repeated each time, exiting the current cycle when one of the conditions is met, completing the sampling of the current target agent model, and determining a training sample of the full-flight domain unsteady aerodynamic agent-reduced model according to the sampling;
8) Generating a large number of design points which are randomly distributed and fully filled in the full flight domain by using a random sampling method; for each design point, obtaining input and output signals of each design point in a training sample by utilizing Kriging interpolation, and determining an unsteady aerodynamic reduced order model by utilizing a discrete space formed by each input and output signal;
9) Converting the unsteady aerodynamic force reduced order model into a state space equation of a continuous space; converting the rigid body power equation into a state equation under continuous space; the state space equation of the unsteady aerodynamic force reduced order model is connected with the state equation of the rigid body dynamic equation in a feedback way, and then the coupling dynamic stability analysis equation of the current aircraft is obtained; solving a characteristic matrix characteristic root of a coupling dynamic stability analysis equation, wherein a real part of the characteristic root represents system damping, and an imaginary part represents system frequency; when all the real parts of the characteristic roots are negative, the aircraft of the design point is stable in motion; when the characteristic root of the positive real part appears, the aircraft movement of the design point is unstable;
10 After the dynamic stability characteristics of the aircraft at each design point are obtained through the steps, the dynamic stability characteristics of the aircraft in the whole flight domain are obtained.
Further, in the step three, the data analysis program is utilized to analyze and judge the risk of the obstacle according to the acquired image and the obstacle information,
detecting an obstacle microscopic image by adopting a pulse coupling neural network model suitable for processing obstacle image information; the obstacle microscopic image is polluted by impulse noise with smaller density and is processed by self-adaptive weighting filtering; the obstacle microscopic image is polluted by impulse noise with high density, and the obstacle dangerous information is obtained after secondary filtering by adopting the mathematical morphology of the introduced double structural elements for keeping edge detail information.
Further, a pulse coupled neural network model adapted to process obstacle image information:
F ij [n]＝S ij ；
U ij [n]＝F ij [n](1+β ij [n]L ij [n])；
θ ij [n]＝θ 0 e -αθ(n-1) ；
wherein beta is ij [n]Is an adaptive link strength coefficient;
S ij 、F ij [n]、L ij [n]、U ij [n]、θi j [n]respectively input image signal, feedback input, link input, internal activity item and dynamic threshold value, N w For the total number of pixels in the selected window W to be processed, delta is an adjustment coefficient, and 1-3 are selected.
Further, when the pulse coupled neural network model detects the obstacle microscopic image, the gray level is S by utilizing the network characteristic ijmax Is activated by the ignition of the pixels, and then carries out the iterative processing of the second pulse coupling neural network to lead the pixel to be between [ Si ] jmax /1+βi j L ij ,Si jmax ]Capturing and activating pixels in between to enable the pixels activated twice to correspond to Y ij The output is 1; then the original noise pollution image is whitened, and the processed image S is processed ij Performing iterative processing as described above, and making the corresponding output Y ij Using the characteristic that the correlation between the image noise pixel and the surrounding pixels is small and the gray scale difference is large, when the excitation of one neuron does not cause the excitation of most of the neurons near the area, the corresponding pixel of the neuron is probably a noise point;
preliminary screening of Y ij The pixel point corresponding to the value of the number of the pixels=0 is a signal point of an obstacle microscopic image and is protected; for Yi j The pixel point with output of 1 is counted in the range of 3*3 template B to output Yi j Number N of=1 for central neighborhood element value 1 Y Judging and classifying: n is more than or equal to 1 Y Less than or equal to 8, as noise point, when N Y =9, and the image pixel is determined.
Another object of the present invention is to provide an autonomous obstacle avoidance system for an aircraft implementing the autonomous obstacle avoidance method for an aircraft of the claims, the autonomous obstacle avoidance system for an aircraft comprising:
the solar power supply module is connected with the central control module and used for converting solar energy into electric energy through the solar cell panel to supply power for the aircraft;
the image acquisition module is connected with the central control module and is used for acquiring an environment image in front of the flight of the aircraft through the optical camera;
the obstacle information detection module is connected with the central control module and is used for carrying out multi-target detection on the obstacle in the front detectable distance through the millimeter wave radar to obtain the position, speed and azimuth information of the obstacle;
the central control module is connected with the solar power supply module, the image acquisition module, the obstacle information detection module, the analysis and judgment module, the instruction generation module, the path planning module, the stability analysis module and the display module and used for controlling the normal work of each module through the central controller;
the analysis and judgment module is connected with the central control module and is used for analyzing and judging the risk of the obstacle according to the acquired image and obstacle information through a data analysis program;
the command generation module is connected with the central control module and is used for generating a route change command according to the judgment result through the command generator;
the path planning module is connected with the central control module and is used for planning the flight path of the aircraft according to the change instruction through the navigation system;
the stability analysis module is connected with the central control module and used for analyzing the stability of the aircraft through an analysis program;
and the display module is connected with the central control module and used for displaying and collecting environmental images, obstacle information, flight paths and stability data through the display.
Another object of the invention is to provide an aircraft.
The invention has the advantages and positive effects that: according to the invention, when other aircrafts except the aircrafts fly in the flight planning height layer, the path planning module judges whether the other aircrafts collide with the aircrafts according to the preset rule, and adjusts the flight planning track when judging that the other aircrafts collide with the aircrafts, so that the flight track for planning safety for the aircrafts can be ensured, and the flight safety is greatly improved; meanwhile, the stability analysis module is used for analyzing the dynamic stability of the whole flight domain multichannel coupling of the aircraft, and the method for solving the characteristic matrix characteristic root in the state space by coupling the unsteady aerodynamic force reduced-order model and the rigid dynamic equation based on the high-efficiency self-adaptive sampling method is provided for predicting the dynamic stability characteristic of the whole flight domain of the aircraft, so that the calculation cost can be greatly reduced while the prediction precision is ensured, qualitative and quantitative research on the dynamic stability characteristic of the whole flight domain of the aircraft is easy to develop, and the design direction with guiding significance on the design of the aircraft is obtained.
In the invention, in the process that the image acquisition module acquires the environment image in front of the aircraft through the optical camera, under the condition that the shielding of multiple moving targets is overlapped, a multi-target tracking algorithm based on particle filtering is adopted, so that the tracking success rate of the moving targets is improved, the tracking success rate of the targets is increased by 39.5 percentage points, and the average time consumption of the algorithm is 0.78s.
In the invention, the obstacle information detection module carries out multi-target detection on the obstacle in the front detectable distance through the millimeter wave radar, and in the process of obtaining the position, speed and azimuth information of the obstacle, in the complex detection background, the echo signal is difficult to avoid being doped with a plurality of clutter components, and the adoption of the LVQ clustering algorithm can eliminate the clutter components from being left in the form of false targets.
In the invention, the analysis and judgment module adopts an improved decision tree-based data analysis and judgment algorithm in the process of analyzing and judging the risk of the obstacle according to the acquired image and obstacle information by the data analysis program, so that a decision tree of the risk degree of the obstacle can be quickly established according to the acquired image and obstacle data information, and a basis is provided for quickly determining the risk of the obstacle.
According to the invention, by improving the pulse coupling neural network, noise points in the obstacle microscopic image are automatically detected without setting a detection threshold value, and the multistage combined filter is utilized to remove noise, so that the noise interference is effectively filtered, and meanwhile, the information such as image edge details and the like is well protected.
The invention has the following effects:
the method utilizes the synchronous pulse issuing characteristic of the pulse coupling neural network to distinguish and position the impulse noise point and the signal pixel point, has higher noise point detection performance compared with the traditional median detection method based on median detection or related improvement, and is compared with other threshold noise point detection methods; the invention does not need to set a detection threshold, has low noise false detection rate and omission rate and higher noise detection precision; meanwhile, relative to other noise iterative detection methods; the method has short detection time and strong automaticity;
at present, no impulse noise detection method is applied to detection of impulse noise of obstacle microscopic images; in the stage of filtering impulse noise of the obstacle microscopic image, the invention firstly carries out classification processing on the image pixels according to the detected noise points and signal points; when the first-stage self-adaptive weighted filtering is utilized, filtering processing is only carried out on detected noise points, and compared with other methods such as median filtering, wiener filtering and the like, the method effectively filters the noise points and simultaneously protects signal point information; in the second stage of morphological filtering, the relevant noise points which are missed in the previous stage of filtering are subjected to supplementary auxiliary filtering, so that noise interference can be effectively filtered while denoising is performed, and information such as image edge details and the like can be well protected;
the method has the advantages of strong subjective visual effect and objective evaluation index, strong denoising capability, high signal-to-noise ratio and good adaptability, and particularly shows greater filtering superiority for obstacle microscopic images polluted by serious noise.
Drawings
Fig. 1 is a flowchart of an autonomous obstacle avoidance method for an aircraft according to an embodiment of the present invention.
Fig. 2 is a block diagram of an autonomous obstacle avoidance system of an aircraft according to an embodiment of the present invention.
In fig. 2: 1. a solar power supply module; 2. an image acquisition module; 3. an obstacle information detection module; 4. a central control module; 5. an analysis and judgment module; 6. an instruction generation module; 7. a path planning module; 8. a stability analysis module; 9. and a display module.
Detailed Description
For a further understanding of the invention, its features and advantages, reference is now made to the following examples, which are included in the detailed description taken in conjunction with the accompanying drawings.
The structure of the present invention will be described in detail with reference to the accompanying drawings.
As shown in fig. 1, the autonomous obstacle avoidance method of the aircraft provided by the invention comprises the following steps:
s101: first, solar panels convert solar energy into electrical energy for powering the aircraft.
S102: and acquiring an environment image in front of the flying of the aircraft, and carrying out multi-target detection on the obstacle in the front detectable distance through the radar to obtain the position, speed and azimuth information of the obstacle.
S103: and analyzing and judging the risk of the obstacle according to the acquired data information, generating a route changing instruction according to the judging result, and planning the flight path of the aircraft.
S104: and acquiring flight state information of the aircraft, and analyzing the stability of the aircraft.
S105: and displaying the acquired environment images, obstacle information, flight paths and stability data through a display.
Step S103, analyzing and judging the risk of the obstacle according to the acquired image and the obstacle information by utilizing a data analysis program,
detecting an obstacle microscopic image by adopting a pulse coupling neural network model suitable for processing obstacle image information; the obstacle microscopic image is polluted by impulse noise with smaller density and is processed by self-adaptive weighting filtering; the obstacle microscopic image is polluted by impulse noise with high density, and the obstacle dangerous information is obtained after secondary filtering by adopting the mathematical morphology of the introduced double structural elements for keeping edge detail information.
Pulse coupled neural network model adapted to process obstacle image information:
F ij [n]＝S ij ；
U ij [n]＝F ij [n](1+β ij [n]L ij [n])；
θ ij [n]＝θ 0 e -αθ(n-1) ；
wherein beta is ij [n]Is an adaptive link strength coefficient;
S ij 、F ij [n]、L ij [n]、U ij [n]、θi j [n]respectively input image signal, feedback input, link input, internal activity item and dynamic threshold value, N w For the total number of pixels in the selected window W to be processed, delta is an adjustment coefficient, and 1-3 are selected.
When the pulse coupling neural network model detects the obstacle microscopic image, the gray level is S by utilizing the network characteristic ijmax Is activated by the ignition of the pixels, and then carries out the iterative processing of the pulse coupling neural network for the second time, which is between [ S ] ijmax /1+β ij L ij ,S ijmax ]Capturing and activating pixels in between to enable the pixels activated twice to correspond to Y ij The output is 1; then the original noise pollution image is whitened, and the processed image S is processed ij Performing iterative processing as described above, and making the corresponding output Y ij By using the characteristic that the correlation between the image noise pixel and the surrounding pixels is small and the gray scale difference is large, =1, when the excitation of one neuron does not cause the excitation of most of the neurons near the region, it is indicated that the corresponding pixel of the neuron may be a noise point；
Preliminary screening of Y ij The pixel point corresponding to the value of the number of the pixels=0 is a signal point of an obstacle microscopic image and is protected; for Y ij The pixel point with output of 1 is counted in the range of 3*3 template B to output Y ij Number N of=1 for central neighborhood element value 1 Y Judging and classifying: n is more than or equal to 1 Y Less than or equal to 8, as noise point, when N Y =9, determined as image pixel points;
the method for realizing noise filtering of the obstacle microscopic image self-adaptive weighting filter comprises the following steps of;
when pulse outputs Y ij =1 and N Y ＝1～8，N Y When the number is 1 in the 3*3 template B, a filtering window M is selected to pollute the image f with noise ij Is the adaptive filtering of (1), the filtering equation is:
wherein x is rs Is the coefficient of the corresponding pixel in the filter window, S rs For filtering the gray value of the corresponding pixel in the window, f ij The output value of the center position of the corresponding window after filtering is as follows:
d in ij Is the median value of pixel gray scale in square filter window M, omega ij The absolute average value of the gray differences between each pixel and the center of the filter window, and max is the maximum value sign;
selecting a filter window M, and selecting a filter window M with the size of M x M, wherein the window size is selected according to the following principle:
the specific method of the mathematical morphology second-stage filtering of the double structural elements comprises the following steps:
if the obstacle microscopic image of the residual impulse noise is f and E is a structural element SE, the expansion has the following relation:
in the middle ofFor the expansion operator, F and G are the definition fields of F and E, respectively, and x-z is the displacement parameter;
the expansion relation is a process of combining all background points contacted with the object into the object to expand the boundary outwards and fill holes in the object;
the above-mentioned Θ is corrosion operator, corrosion is to eliminate boundary point, boundary is contracted inwards, on the basis of corrosion expansion, and then combining morphological opening and closing operation:
as shown in fig. 2, an autonomous obstacle avoidance system for an aircraft according to an embodiment of the present invention includes: the system comprises a solar power supply module 1, an image acquisition module 2, an obstacle information detection module 3, a central control module 4, an analysis and judgment module 5, an instruction generation module 6, a path planning module 7, a stability analysis module 8 and a display module 9.
The solar power supply module 1 is connected with the central control module 4 and is used for converting solar energy into electric energy through a solar panel to supply power for the aircraft;
the image acquisition module 2 is connected with the central control module 4 and is used for acquiring an environment image in front of the flight of the aircraft through the optical camera;
the obstacle information detection module 3 is connected with the central control module 4 and is used for carrying out multi-target detection on the obstacle in the front detectable distance through the millimeter wave radar to obtain the position, speed and azimuth information of the obstacle;
the central control module 4 is connected with the solar power supply module 1, the image acquisition module 2, the obstacle information detection module 3, the analysis and judgment module 5, the instruction generation module 6, the path planning module 7, the stability analysis module 8 and the display module 9 and is used for controlling the normal work of each module through the central controller;
the analysis and judgment module 5 is connected with the central control module 4 and is used for analyzing and judging the risk of the obstacle according to the acquired image and obstacle information through a data analysis program;
the instruction generation module 6 is connected with the central control module 4 and is used for generating a route change instruction according to the judgment result through the instruction generator;
the path planning module 7 is connected with the central control module 4 and is used for planning the flight path of the aircraft according to the change instruction through the navigation system;
the stability analysis module 8 is connected with the central control module 4 and is used for analyzing the stability of the aircraft through an analysis program;
and the display module 9 is connected with the central control module 4 and is used for displaying and collecting environmental images, obstacle information, flight paths and stability data through a display.
In the process that the image acquisition module 2 acquires the environmental image in front of the aircraft through the optical camera, in order to improve the tracking success rate of the moving target under the condition that the shielding of the multiple moving targets is overlapped, a multi-target tracking algorithm based on particle filtering is adopted, and the method specifically comprises the following steps:
starting M threads for M targets, tracking in parallel, and improving the speed;
step two, for each tracking target, determining the range of the moving target according to the corresponding pre-algorithm at t=0, initializing a particle sample set according to the particle filtering algorithm within the rangeThe weight w 'of all particles at the time t=0' 0 Let 1/N, i=1, 2, …, N; assume that the target is located +.>
Step three, let t=t+1, according to formula
Calculate->
Wherein g (x) = -k' (x);
according to
Wherein the method comprises the steps ofAnd->Is a candidate target;
calculating a Pasteur coefficient ρ representing the similarity of the tracked position to the target, taking G 1 ＝0.85，G 2 ＝0.35；
When ρ is greater than or equal to G 1 When the tracking is normal, executing the step six; when G 2 ρ＜G 1 When the shielding occurs, the shielding is partially performed, and the fourth step is executed; when ρ is less than or equal to G 2 When the method is used, the occurrence of shielding is indicated, most of the steps are performed by shieldingStep five;
tracking according to the segmented image module to obtain againReturning to the third step to recalculate the Pasteur coefficient;
step six, byNew location, reassignment of particle location, update to get new particle set +.>New weight->The calculation is that
Wherein lambda is a control parameter, a weight coefficient is set according to the distance between the tracking target and the pasteurization distance, the weight is higher when the pasteurization distance is closer, the weight is lower when the pasteurization distance is farther, and the weight is normalized
Step seven, judging whether all frames of the video are processed, if not, processing the next frame of image, and executing the step three; if yes, executing the step eight; the algorithm ends and the processing thread is released.
The obstacle information detection module 3 performs multi-target detection on obstacles in a front detectable distance through a millimeter wave radar, and in the process of obtaining position, speed and azimuth information of the obstacles, in a complex detection background, the echo signals are difficult to avoid being doped with a plurality of clutter components, and in order to eliminate the clutter components, the LVQ clustering algorithm is adopted, wherein the LVQ clustering algorithm specifically comprises the following steps:
step one, sample data collected for specific targets (aircrafts and ships) under the condition of real echo, wherein the information of distance, azimuth, elevation and the like is basically known; setting the true and false labels of the part of known targets in the sample as true, and setting all labels of other targets as false;
step two, randomly extracting a prototype vector from the sample set with the original label being true/false respectively for initializing the prototype vectorAnd->And will->And->Cluster labels of (2) are set to true/false, respectively;
step three, randomly extracting a group of sample vectors from all samplesCalculate it and two prototype vectors +.>Andfor prototype vectors with smaller distance values +.>According to the extracted sample->Category label y of (2) j And->Is the label of (2)If not, respectively use the following
p′＝p i ·+α·(x j -p j ·)；
p′＝p i ·-α·(x j -p i ·)；
For prototype vectorUpdating, wherein alpha epsilon (0, 1) is the learning rate;
step four, outputting prototype vector representing real and false target trace characteristics when the maximum iteration times or the condition that the prototype vector is updated little or not is satisfiedAnd->
In the process of analyzing and judging the risk of the obstacle according to the collected image and obstacle information by the data analysis program, the analysis and judgment module 5 provides a basis for quickly determining the risk of the obstacle in order to quickly establish a decision tree of the risk degree of the obstacle according to the collected image and obstacle data information, adopts an improved decision tree-based data analysis and judgment algorithm, and comprises the following specific steps:
step one, a training set sample T is established, and the training set sample T is returned when meeting the expansion stopping condition;
step two, scanning an attribute list for discrete attributes, and updating a counting matrix to determine the optimal split subset of the attributes;
dividing the continuous attribute into q intervals by adopting an equal-width histogram method, and establishing a partition histogram list;
step four, calculating the Gini value at the boundary of each pure interval;
step five, pruning the interval by using a TESTCASE algorithm;
step six, judging whether the tuples segmented by partial segmentation points in the candidate interval are the same, and deleting the candidate points with the same segmentation tuple;
step seven, finding out the optimal splitting point of the whole attribute; comparing the optimal splits of the respective attributes, selecting an optimal split point to divide T into T 1 And T 2 ；
Step eight, recursively pair T 1 And T 2 A decision tree is generated.
The path planning module 7 planning method provided by the invention comprises the following steps:
(1) Receiving a route change instruction, and generating a flight planning track for the aircraft, wherein the flight planning track comprises a flight planning route and a flight planning height layer;
(2) When other aircrafts except the aircrafts fly in the flight planning altitude layer, judging whether the other aircrafts collide with the aircrafts according to a preset rule;
(3) And when judging that the other aircrafts collide with the aircrafts, adjusting the flight planning track until judging that the other aircrafts do not collide with the aircrafts according to the preset rule.
The method for judging whether the other aircrafts collide with the aircrafts according to the preset rule provided by the invention comprises the following steps:
acquiring flight trajectories and flight speeds of the other aircrafts;
when the flight planning track and the flight track of the other aircrafts have a superposition area, calculating the time point interval of the other aircrafts and the aircrafts reaching the superposition area according to the flight track and the flight speed of the other aircrafts and the flight planning track and the flight speed of the aircrafts;
and when the time point interval is smaller than or equal to a preset time interval, judging that the other aircrafts collide with the aircrafts.
The invention provides a method for adjusting the flight planning track of an aircraft, which comprises the following steps:
adjusting the flight plan trajectory by adjusting a flight plan altitude layer and/or a flight plan route of the aircraft at least at the overlap region; the adjusted flight planning altitude of the aircraft is located between a lower limit altitude and an upper limit altitude of the aircraft.
The analysis method of the stability analysis module 8 provided by the invention comprises the following steps:
1) Setting flight parameters of an aircraft;
2) Aiming at aerodynamic characteristics of an aircraft, determining target precision requirements and an upper limit of calculated amount, and selecting a predetermined number of initial sample points in a full flight domain range;
3) At an initial sample point x (i), a time sequence y (x (i), k) of unsteady aerodynamic force and moment coefficients of forced movement of the aircraft is obtained by using CFD calculation for a given set of movement training signal sequences u (x (i), k), and then a Kriging agent model is built as an initial target agent model by taking u (x (i), k) and y (x (i), k) as sample data;
4) Constructing a reference agent model, and evaluating an initial sample point to determine the current sampling precision;
5) Generating new candidate sampling points by using a test design method; calculating and evaluating all candidate sampling points through the target agent model and the reference agent model respectively to obtain precision change values generated after the candidate sampling points are added into the target agent model and the reference agent model;
6) Selecting candidate sampling points meeting the current sampling precision requirement in the precision change value as sample points added into the target agent model and the reference agent model for training through a self-adaptive sampling criterion;
7) Repeating the steps 4) to 5), judging whether the precision of the current target agent model meets the standard and whether the calculated amount reaches the upper limit or not when the current target agent model is repeated each time, exiting the current cycle when one of the conditions is met, completing the sampling of the current target agent model, and determining a training sample of the full-flight domain unsteady aerodynamic agent-reduced model according to the sampling;
8) Generating a large number of design points which are randomly distributed and fully filled in the full flight domain by using a random sampling method; for each design point, obtaining input and output signals of each design point in a training sample by utilizing Kriging interpolation, and determining an unsteady aerodynamic reduced order model by utilizing a discrete space formed by each input and output signal;
9) Converting the unsteady aerodynamic force reduced order model into a state space equation of a continuous space; converting the rigid body power equation into a state equation under continuous space; the state space equation of the unsteady aerodynamic force reduced order model is connected with the state equation of the rigid body dynamic equation in a feedback way, and then the coupling dynamic stability analysis equation of the current aircraft is obtained; solving a characteristic matrix characteristic root of a coupling dynamic stability analysis equation, wherein a real part of the characteristic root represents system damping, and an imaginary part represents system frequency; when all the real parts of the characteristic roots are negative, the aircraft of the design point is stable in motion; when the characteristic root of the positive real part appears, the aircraft movement of the design point is unstable;
10 After the dynamic stability characteristics of the aircraft at each design point are obtained through the steps, the dynamic stability characteristics of the aircraft in the whole flight domain can be obtained.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the invention in any way, but any simple modification, equivalent variation and modification of the above embodiments according to the technical principles of the present invention are within the scope of the technical solutions of the present invention.
Claims (1)
1. An autonomous obstacle avoidance system for an aircraft, comprising: the system comprises a solar power supply module, an image acquisition module, an obstacle information detection module, a central control module, an analysis and judgment module, an instruction generation module, a path planning module, a stability analysis module and a display module;
the solar power supply module is connected with the central control module and used for converting solar energy into electric energy through the solar cell panel to supply power for the aircraft;
the image acquisition module is connected with the central control module and is used for acquiring an environment image in front of the flight of the aircraft through the optical camera;
the obstacle information detection module is connected with the central control module and is used for carrying out multi-target detection on the obstacle in the front detectable distance through the millimeter wave radar to obtain the position, speed and azimuth information of the obstacle;
the central control module is connected with the solar power supply module, the image acquisition module, the obstacle information detection module, the analysis and judgment module, the instruction generation module, the path planning module, the stability analysis module and the display module and used for controlling the normal work of each module through the central controller;
the analysis and judgment module is connected with the central control module and is used for analyzing and judging the risk of the obstacle according to the acquired image and obstacle information through a data analysis program;
the command generation module is connected with the central control module and is used for generating a route change command according to the judgment result through the command generator;
the path planning module is connected with the central control module and is used for planning the flight path of the aircraft according to the change instruction through the navigation system;
the stability analysis module is connected with the central control module and used for analyzing the stability of the aircraft through an analysis program;
the display module is connected with the central control module and used for displaying and collecting environmental images, obstacle information, flight paths and stability data through a display;
in the process that the image acquisition module acquires an environment image in front of the aircraft through the optical camera, a multi-target tracking algorithm based on particle filtering is adopted, and the multi-target tracking algorithm is used for improving the tracking success rate of a moving target under the condition that shielding of multiple moving targets is overlapped, and specifically comprises the following steps:
starting M threads for M targets, tracking in parallel, and improving the speed;
step two, for each tracking target, determining the range of the moving target according to a corresponding pre-algorithm at t=0, and initializing a particle sample set according to a particle filtering algorithm within the range:weights of all particles at time t=0 +.>Let 1/N, i=1, 2, …, N; assume that the target is located +.>；
Step three, let t=t+1, calculate；
According to；
Wherein the method comprises the steps ofAnd->Is a candidate target;
calculating a pasteurization coefficient ρ, which represents the similarity of the tracked position to the target, taking g1=0.85, g2=0.35;
when rho is more than or equal to G1, indicating that tracking is normal, and executing the step six; when G2 rho is smaller than G1, the shielding is shown, but part of the shielding is realized, and the fourth step is executed; when rho is less than or equal to G2, indicating that shielding occurs, wherein most of the shielding occurs, and executing the fifth step;
tracking according to the segmented image module to obtain againReturning to the third step to recalculate the Pasteur coefficient;
step six, byNew location, reassignment of particle location, update to get new particle set +.>New weight->The calculation is that
；
Wherein lambda is a control parameter, a weight coefficient is set according to the distance between the tracking target and the pasteurization distance, the weight is higher when the pasteurization distance is closer, the weight is lower when the pasteurization distance is farther, and the weight is normalized；
Step seven, judging whether all frames of the video are processed, if not, processing the next frame of image, and executing the step three; if yes, executing the step eight; ending the algorithm, and releasing the processing thread;
the obstacle information detection module carries out multi-target detection on obstacles in a front detectable distance through a millimeter wave radar, and in the process of obtaining position, speed and azimuth information of the obstacles, in a complex detection background, the echo signals are difficult to avoid being doped with a plurality of clutter components, and the clutter components for elimination are left in a false target mode, and an LVQ clustering algorithm is adopted, and specifically comprises the following steps:
step one, sample data collected for a specific target under a real echo condition, wherein the information of the distance, the azimuth and the elevation angle of the sample data is known; setting the true and false labels of the part of known targets in the sample as true, and setting all labels of other targets as false;
step two, randomly extracting a prototype vector from the sample set with the original label being true/false respectively for initializing the prototype vectorAndand will->And->Cluster labels of (2) are set to true/false, respectively;
step three, randomly extracting a group of sample vectors from all samplesCalculate it and two prototype vectors +.>And->For prototype vectors with smaller distance values +.>According to the extracted sample->Category label y of (2) j And->If the labels of (2) are identical, the following formulas are used respectively:
or->，
For prototype vectorUpdating, wherein alpha epsilon (0, 1) is the learning rate;
step four, outputting prototype vector representing real and false target trace characteristics when the maximum iteration times or the condition that the prototype vector is updated little or not is satisfiedAnd->；
In the process of analyzing and judging the risk of the obstacle according to the acquired image and obstacle information by the data analysis program, the analysis and judgment module provides a basis for quickly determining the risk of the obstacle in order to quickly establish a decision tree of the risk degree of the obstacle according to the acquired image and obstacle data information, and adopts an improved decision tree-based data analysis and judgment algorithm, which comprises the following specific steps:
step one, a training set sample T is established, and the training set sample T is returned when meeting the expansion stopping condition;
step two, scanning an attribute list for discrete attributes, and updating a counting matrix to determine the optimal split subset of the attributes;
dividing the continuous attribute into q intervals by adopting an equal-width histogram method, and establishing a partition histogram list;
step four, calculating the Gini value at the boundary of each pure interval;
step five, pruning the interval by using a TESTCASE algorithm;
step six, judging whether the tuples segmented by partial segmentation points in the candidate interval are the same, and deleting the candidate points with the same segmentation tuple;
step seven, finding out the optimal splitting point of the whole attribute; comparing the optimal splits of the respective attributes, selecting an optimal split point to divide T into T 1 And T 2 ；
Step eight, recursively pair T 1 And T 2 Generating a decision tree;
the planning method of the path planning module comprises the following steps:
(1) Receiving a route change instruction, and generating a flight planning track for the aircraft, wherein the flight planning track comprises a flight planning route and a flight planning height layer;
(2) When other aircrafts except the aircrafts fly in the flight planning altitude layer, judging whether the other aircrafts collide with the aircrafts according to a preset rule;
(3) When the other aircrafts are judged to collide with the aircrafts, the flight planning track is adjusted until the other aircrafts are judged not to collide with the aircrafts according to the preset rules;
judging whether other aircrafts collide with the aircrafts according to a preset rule, including:
acquiring flight trajectories and flight speeds of the other aircrafts;
when the flight planning track and the flight track of the other aircrafts have a superposition area, calculating the time point interval of the other aircrafts and the aircrafts reaching the superposition area according to the flight track and the flight speed of the other aircrafts and the flight planning track and the flight speed of the aircrafts;
when the time point interval is smaller than or equal to a preset time interval, judging that the other aircrafts collide with the aircrafts;
adjusting a flight planning trajectory of the aircraft, comprising:
adjusting the flight plan trajectory by adjusting a flight plan altitude layer and/or a flight plan route of the aircraft at least at the overlap region; the adjusted flight planning height of the aircraft is located between the lower limit flight height and the upper limit flight height of the aircraft;
the analysis method of the stability analysis module comprises the following steps:
1) Setting flight parameters of an aircraft;
2) Aiming at aerodynamic characteristics of an aircraft, determining target precision requirements and an upper limit of calculated amount, and selecting a predetermined number of initial sample points in a full flight domain range;
3) At an initial sample point x (i), a time sequence y (x (i), k) of unsteady aerodynamic force and moment coefficients of forced movement of the aircraft is obtained by using CFD calculation for a given set of movement training signal sequences u (x (i), k), and then a Kriging agent model is built as an initial target agent model by taking u (x (i), k) and y (x (i), k) as sample data;
4) Constructing a reference agent model, and evaluating an initial sample point to determine the current sampling precision;
5) Generating new candidate sampling points by using a test design method; calculating and evaluating all candidate sampling points through the target agent model and the reference agent model respectively to obtain precision change values generated after the candidate sampling points are added into the target agent model and the reference agent model;
6) Selecting candidate sampling points meeting the current sampling precision requirement in the precision change value as sample points added into the target agent model and the reference agent model for training through a self-adaptive sampling criterion;
7) Repeating the steps 4) to 5), judging whether the precision of the current target agent model meets the standard and whether the calculated amount reaches the upper limit or not when the current target agent model is repeated each time, exiting the current cycle when one of the conditions is met, completing the sampling of the current target agent model, and determining a training sample of the full-flight domain unsteady aerodynamic agent-reduced model according to the sampling;
8) Generating a large number of design points which are randomly distributed and fully filled in the full flight domain by using a random sampling method; for each design point, obtaining input and output signals of each design point in a training sample by utilizing Kriging interpolation, and determining an unsteady aerodynamic reduced order model by utilizing a discrete space formed by each input and output signal;
9) Converting the unsteady aerodynamic force reduced order model into a state space equation of a continuous space; converting the rigid body power equation into a state equation under continuous space; the state space equation of the unsteady aerodynamic force reduced order model is connected with the state equation of the rigid body dynamic equation in a feedback way, and then the coupling dynamic stability analysis equation of the current aircraft is obtained; solving a characteristic matrix characteristic root of a coupling dynamic stability analysis equation, wherein a real part of the characteristic root represents system damping, and an imaginary part represents system frequency; when all the real parts of the characteristic roots are negative, the aircraft of the design point is stable in motion; when the characteristic root of the positive real part appears, the aircraft movement of the design point is unstable;
10 After the dynamic stability characteristics of the aircraft at each design point are obtained through the steps, the dynamic stability characteristics of the aircraft in the whole flight domain can be obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910132801.6A CN109634309B (en) | 2019-02-21 | 2019-02-21 | Autonomous obstacle avoidance system and method for aircraft and aircraft |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910132801.6A CN109634309B (en) | 2019-02-21 | 2019-02-21 | Autonomous obstacle avoidance system and method for aircraft and aircraft |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109634309A CN109634309A (en) | 2019-04-16 |
| CN109634309B true CN109634309B (en) | 2024-03-26 |
Family
ID=66065782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910132801.6A Active CN109634309B (en) | 2019-02-21 | 2019-02-21 | Autonomous obstacle avoidance system and method for aircraft and aircraft |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109634309B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110299030B (en) * | 2019-06-28 | 2021-11-19 | 汉王科技股份有限公司 | Handheld terminal, aircraft, airspace measurement method and control method of aircraft |
| CN111175730A (en) * | 2020-01-08 | 2020-05-19 | 中国石油大学(华东) | Millimeter wave radar target trace condensing method for unmanned ship |
| CN112069889B (en) * | 2020-07-31 | 2021-08-03 | 北京信息科技大学 | Civil aircraft trajectory prediction method, electronic device and storage medium |
| CN112356761B (en) * | 2020-09-22 | 2022-06-14 | 成都飞机工业（集团）有限责任公司 | Aircraft transportation method based on whole machine erection |
| CN112306092A (en) * | 2020-10-29 | 2021-02-02 | 吉林化工学院 | Unmanned aerial vehicle system of patrolling and examining |
| CN113311857A (en) * | 2021-04-29 | 2021-08-27 | 重庆交通大学 | Environment sensing and obstacle avoidance system and method based on unmanned aerial vehicle |
| CN113467495B (en) * | 2021-07-07 | 2023-08-11 | 中国人民解放军火箭军工程大学 | Aircraft rapid trajectory optimization method meeting strict time and position constraints |
| CN113467504B (en) * | 2021-07-26 | 2023-06-02 | 广东电网有限责任公司 | Method, system, equipment and storage medium for controlling flight stability of aircraft |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104732500A (en) * | 2015-04-10 | 2015-06-24 | 天水师范学院 | Traditional Chinese medicinal material microscopic image noise filtering system and method adopting pulse coupling neural network |
| CN107272731A (en) * | 2017-06-05 | 2017-10-20 | 陈金良 | The automatic anti-collision system of unmanned plane |
| CN107831777A (en) * | 2017-09-26 | 2018-03-23 | 中国科学院长春光学精密机械与物理研究所 | A kind of aircraft automatic obstacle avoiding system, method and aircraft |
| CN108121856A (en) * | 2017-12-06 | 2018-06-05 | 中国科学院力学研究所 | A kind of full flight domain aerocraft dynamic stability analysis method |
| WO2019015158A1 (en) * | 2017-07-21 | 2019-01-24 | 歌尔科技有限公司 | Obstacle avoidance method for unmanned aerial vehicle, and unmanned aerial vehicle |
-
2019
- 2019-02-21 CN CN201910132801.6A patent/CN109634309B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104732500A (en) * | 2015-04-10 | 2015-06-24 | 天水师范学院 | Traditional Chinese medicinal material microscopic image noise filtering system and method adopting pulse coupling neural network |
| CN107272731A (en) * | 2017-06-05 | 2017-10-20 | 陈金良 | The automatic anti-collision system of unmanned plane |
| WO2019015158A1 (en) * | 2017-07-21 | 2019-01-24 | 歌尔科技有限公司 | Obstacle avoidance method for unmanned aerial vehicle, and unmanned aerial vehicle |
| CN107831777A (en) * | 2017-09-26 | 2018-03-23 | 中国科学院长春光学精密机械与物理研究所 | A kind of aircraft automatic obstacle avoiding system, method and aircraft |
| CN108121856A (en) * | 2017-12-06 | 2018-06-05 | 中国科学院力学研究所 | A kind of full flight domain aerocraft dynamic stability analysis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109634309A (en) | 2019-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109634309B (en) | Autonomous obstacle avoidance system and method for aircraft and aircraft | |
| US20220197281A1 (en) | Intelligent decision-making method and system for unmanned surface vehicle | |
| CN107909082B (en) | Sonar image target identification method based on deep learning technology | |
| CN114048889B (en) | Aircraft trajectory prediction method based on long-term and short-term memory network | |
| US4906940A (en) | Process and apparatus for the automatic detection and extraction of features in images and displays | |
| CN112434643A (en) | Classification and identification method for low-slow small targets | |
| CN110018453B (en) | Intelligent model identification method based on aircraft track characteristics | |
| CN108320051B (en) | Mobile robot dynamic collision avoidance planning method based on GRU network model | |
| CN111123257A (en) | Radar moving target multi-frame joint detection method based on graph space-time network | |
| Shi et al. | Objects detection of UAV for anti-UAV based on YOLOv4 | |
| CN111539422B (en) | Flight target cooperative identification method based on fast RCNN | |
| CN111027627A (en) | Vibration information terrain classification and identification method based on multilayer perceptron | |
| CN113808174A (en) | Radar small target tracking method based on full convolution network and Kalman filtering | |
| CN111445498A (en) | Target tracking method adopting Bi-L STM neural network | |
| Fan et al. | Improvement of object detection based on faster R-CNN and YOLO | |
| CN113064133B (en) | Sea surface small target feature detection method based on time-frequency domain depth network | |
| CN112556682B (en) | Automatic detection algorithm for underwater composite sensor target | |
| CN113589272A (en) | Automatic generation method for target tracking equipment on-duty log | |
| CN108459614B (en) | UUV real-time collision avoidance planning method based on CW-RNN network | |
| Chen et al. | Sparse Gaussian process regression based ground segmentation for autonomous land vehicles | |
| CN115909086A (en) | SAR target detection and identification method based on multistage enhanced network | |
| CN116304966A (en) | Track association method based on multi-source data fusion | |
| Liu et al. | The development of a UAV target tracking system based on YOLOv3-tiny object detection algorithm | |
| Yang et al. | Study on the trajectory positioning and prediction framework of unmanned aerial vehicle based on long short-term memory neural network | |
| Klaproth et al. | Investigation of Mission Performance Sensitivities for Unmanned Aerial Vehicles Based on Parametric Similarity Assessment of Mission Requirements |
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 |