CN109978045A - A kind of method for tracking target, device and unmanned plane - Google Patents
A kind of method for tracking target, device and unmanned plane Download PDFInfo
- Publication number
- CN109978045A CN109978045A CN201910213970.2A CN201910213970A CN109978045A CN 109978045 A CN109978045 A CN 109978045A CN 201910213970 A CN201910213970 A CN 201910213970A CN 109978045 A CN109978045 A CN 109978045A
- Authority
- CN
- China
- Prior art keywords
- target
- detected
- tracker
- target frame
- frame
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
Abstract
The embodiment of the invention discloses a kind of method for tracking target, device and unmanned planes, which comprises obtains image to be detected;The target in described image to be detected is detected using tracker;Judge the target whether is detected in image to be detected, if the target is not detected in image to be detected, image to be detected is then inputted into the goal-selling detection model based on deep learning, to obtain at least one candidate target frame and the corresponding classification of the candidate target frame;A candidate target frame is selected from least one candidate target frame according to the classification of target and color characteristic;Tracker is updated based on the candidate target frame chosen.When that can not detect the target in image to be detected using tracker, the tracker is updated using the candidate target frame corresponding with target that target detection model obtains.The detectability of tracker can be improved, reduce BREAK TRACK rate.
Description
Technical field
The present embodiments relate to unmanned vehicle technical field, in particular to a kind of method for tracking target, device and nothing
It is man-machine.
Background technique
It carries out intelligent-tracking to moving target using unmanned plane to be widely used, intelligent-tracking can be used for runaway convict and chase after
Track, abnormal object behavioural analysis etc..Currently, filtering algorithm is mostly used to track target, tracking velocity is fast.
In realizing process of the present invention, at least there are the following problems in the related technology for inventor's discovery:
Method for tracking target based on filtering algorithm is blocked or the occasion of target deformation, when especially long in target
Between track target when be easily lost target.
Summary of the invention
The purpose of the embodiment of the present invention is that providing a kind of method for tracking target, device and unmanned plane, target following can be reduced
Loss Rate.
In a first aspect, the embodiment of the invention provides a kind of method for tracking target, which comprises
Obtain image to be detected;
The target in described image to be detected is detected using tracker, wherein spy of the tracker based on the target
Sign training obtains;
Judge the target whether is detected in described image to be detected;
If the target is not detected in described image to be detected, the input of described image to be detected is based on depth
The goal-selling detection model of study, to obtain at least one candidate target frame and the corresponding classification of the candidate target frame;
Candidate target frame is selected from least one described candidate target frame according to the classification of the target and color characteristic;
Based on the candidate target frame re -training trace model chosen, the tracker is updated.
In wherein some embodiments, whether the judgement detects the target in described image to be detected, comprising:
Judge to detect whether the maximum response that described image to be detected obtains is not more than default sound using the tracker
Answer threshold value;
If so, the target is not detected in determination in described image to be detected.
In wherein some embodiments, at least one is candidate from described for the classification and color characteristic according to the target
A candidate target frame is selected in target frame, comprising:
Candidate target frame identical with the target category is selected from least one described candidate target frame;
The color characteristic phase of color characteristic and the target is selected from candidate target frame identical with the target category
Candidate target frame that is maximum like degree and being greater than default similarity threshold.
In wherein some embodiments, whether the judgement detects the target in the feature detection image, packet
It includes:
Judge to detect whether the maximum response that described image to be detected obtains is greater than default response using the tracker
Threshold value;
If so, determination detects the target in described image to be detected.
In wherein some embodiments, this method further include:
Position using the corresponding position of the maximum response as the target in described image to be detected.
In wherein some embodiments, the clarification of objective includes the initial target frame of the target, then, in the benefit
Before detecting the target in described image to be detected with tracker, the method also includes:
Obtain initial target frame;
Based on the initial target frame training trace model, the tracker is obtained.
In wherein some embodiments, the color characteristic includes Color Statistical histogram.
In wherein some embodiments, the method also includes:
The Color Statistical histogram of the target is obtained based on the initial target frame.
In wherein some embodiments, the method also includes:
If it is similar not obtain and color characteristic identical as the target category from least one described candidate target frame
Degree is greater than the candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtains new image to be detected base
It is detected again in goal-selling detection model;
If obtaining and color characteristic maximum phase identical as the target category from least one described candidate target frame
It is greater than the candidate target frame of default similarity threshold like degree, then resets when will be pre-designed;
If numerical value when described pre-designed reaches preset threshold, initial target frame is reacquired, and based on described first
Beginning target frame updates the tracker.
In wherein some embodiments, the tracker is the tracker based on core correlation filtering.
In wherein some embodiments, the goal-selling detection model is the target detection model based on SSD algorithm.
Second aspect, the embodiment of the invention provides a kind of target tracker, described device includes:
Image collection module, for obtaining image to be detected;
Tracker detection module, for detecting the target in described image to be detected using tracker, wherein the tracking
Device is based on clarification of objective training and obtains;
Judgment module, for judging whether detect the target in described image to be detected;
Module of target detection, if for the target to be not detected in described image to be detected, it will be described to be checked
Altimetric image inputs the goal-selling detection model based on deep learning, to obtain at least one candidate target frame and the candidate
The corresponding classification of target frame;
Candidate target frame selecting module, at least one to be candidate from described according to the classification of the target and color characteristic
Candidate target frame is selected in target frame;
First tracker update module, for based on the candidate target frame re -training trace model chosen, described in update
Tracker.
In wherein some embodiments, the judgment module is specifically used for:
Judge to detect whether the maximum response that described image to be detected obtains is not more than default sound using the tracker
Answer threshold value;
If so, the target is not detected in determination in described image to be detected.
In wherein some embodiments, the candidate target frame selecting module is specifically used for:
Candidate target frame identical with the target category is selected from least one described candidate target frame;
The color characteristic phase of color characteristic and the target is selected from candidate target frame identical with the target category
Candidate target frame that is maximum like degree and being greater than default similarity threshold.
In wherein some embodiments, the judgment module also particularly useful for:
Judge to detect whether the maximum response that described image to be detected obtains is greater than default response using the tracker
Threshold value;
If so, determination detects the target in described image to be detected.
In wherein some embodiments, described device further include:
Target position determining module is used for using the corresponding position of the maximum response as the target described to be checked
Position in altimetric image.
In wherein some embodiments, the clarification of objective includes the initial target frame of the target;
Described device further includes tracker training module, for detecting in described image to be detected using the tracker
Target before:
Obtain initial target frame;
Based on the initial target frame training trace model, the tracker is obtained.
In wherein some embodiments, the color characteristic includes Color Statistical histogram.
In wherein some embodiments, described device further include:
Color of object feature obtains module, for obtaining the Color Statistical histogram of the target based on the initial target frame
Figure.
In wherein some embodiments, described device further includes the second tracker update module, is used for:
If it is similar not obtain and color characteristic identical as the target category from least one described candidate target frame
Degree is greater than the candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtains new image to be detected base
It is detected again in goal-selling detection model;
If obtaining and color characteristic maximum phase identical as the target category from least one described candidate target frame
It is greater than the candidate target frame of default similarity threshold like degree, then resets when will be pre-designed;
If numerical value when described pre-designed reaches preset threshold, initial target frame is reacquired, and based on described first
Beginning target frame updates the tracker.
In wherein some embodiments, the tracker is the tracker based on core correlation filtering.
In wherein some embodiments, the goal-selling detection model is the target detection model based on SSD algorithm.
The third aspect, the embodiment of the invention provides a kind of unmanned plane, the unmanned plane include fuselage, with the fuselage phase
Horn even, set on the horn dynamical system, be set to the fuselage photographic device and tracking chip, the camera shooting fill
It sets and is electrically connected with the tracking chip, wherein the photographic device is for obtaining image to be detected, the tracking chip packet
It includes:
At least one processor;And
The memory being connect at least one described processor communication;Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
A processor executes, so that at least one described processor is able to carry out above-mentioned method.
Fourth aspect, the embodiment of the invention provides a kind of non-volatile computer readable storage medium storing program for executing, which is characterized in that
The computer-readable recording medium storage has computer executable instructions, when the computer executable instructions are held by unmanned plane
When row, the unmanned plane is made to execute above-mentioned method.
Method for tracking target, device and the unmanned plane of the embodiment of the present invention, first pass through be obtained ahead of time tracker detection to
Target in detection image detects mould by goal-selling if not detecting the target in described image to be detected
Type detects described image to be detected, obtains at least one candidate target frame and the candidate target in image to be detected
The corresponding classification of frame.Then one is selected from least one described candidate target frame according to the color characteristic of the target and classification
A candidate target frame, re -training trace model obtain new tracker.I.e. when can not detect mapping to be checked using tracker
When target as in, the candidate target frame corresponding with target for utilizing target detection model to obtain updates the tracker.It can mention
The detectability of high tracker reduces BREAK TRACK rate.
Detailed description of the invention
One or more embodiments are illustrated by the picture in corresponding attached drawing, these exemplary theorys
The bright restriction not constituted to embodiment, the element in attached drawing with same reference numbers label are expressed as similar element, remove
Non- to have special statement, composition does not limit the figure in attached drawing.
Fig. 1 is the application scenarios schematic diagram of method for tracking target of the embodiment of the present invention and device;
Fig. 2 is the structural schematic diagram of one embodiment of unmanned plane of the present invention;
Fig. 3 is the schematic diagram of target frame in the embodiment of the present invention;
Fig. 4 is the structural schematic diagram of one embodiment of unmanned plane of the present invention;
Fig. 5 is the flow diagram of one embodiment of method for tracking target of the present invention;
Fig. 6 a is the schematic diagram of training tracker model in the embodiment of the present invention;
Fig. 6 b is the schematic diagram detected using tracker to image to be detected in the embodiment of the present invention;
Fig. 7 is the structural schematic diagram of one embodiment of target tracker of the present invention;
Fig. 8 is the structural schematic diagram of one embodiment of target tracker of the present invention;
Fig. 9 is the hardware structural diagram that chip or controller are tracked in one embodiment of unmanned plane of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Method for tracking target, device and unmanned plane provided in an embodiment of the present invention can be applied to application as shown in Figure 1
Scene please refers to Fig. 1, and in the application scenes of unmanned plane, including unmanned plane 100 and target 200, unmanned plane 100 are used for
Track target 200.Wherein, unmanned plane 100 can include fixed-wing unmanned vehicle and rotary wings for suitable unmanned vehicle
Unmanned vehicle, such as the aircraft of helicopter, quadrotor machine and rotor and/or rotor configuration with other quantity.Nobody
Machine 100 can also be other loose impediments, such as manned vehicle, model plane, unmanned airship and unmanned fire balloon etc..Target
200 can be any suitable removable or irremovable object, including the vehicles, people, animal, building, mountains and rivers river
Deng.
Wherein, in some embodiments, (part-structure of unmanned plane is illustrated only in figure) referring to figure 2., unmanned plane 100
Including fuselage 10, the horn being connected with the fuselage 10, the dynamical system set on horn and the control system set on fuselage 10.It is dynamic
Force system is used to provide thrust, the lift etc. of the flight of unmanned plane 100, and control system is the nervous centralis of unmanned plane 100, be can wrap
Include multiple functional elements, for example, flight control system, tracking system, path planning system and other with specific function be
System.
Wherein, tracking system be used to obtaining track the position of target, tracking range (i.e. unmanned plane 100 away from target 200 away from
From) etc..Flight control system includes various kinds of sensors (such as gyroscope, accelerometer etc.), and flight control system is used to control flying for unmanned plane
Row posture etc..Path planning system is used to plan the flight path of unmanned plane based on the position of tracking target, and indicates
Flight control system controls the flight attitude of unmanned plane so that unmanned plane flies by specified path.
Wherein, tracking system includes photographic device 20 and tracking chip 30, and photographic device 20 and tracking chip 30 electrically connect
It connects, photographic device 20 obtains image to be detected for shooting, and tracking chip 30 determines institute for obtaining described image to be detected
Position of the target in image to be detected is stated, to obtain the actual position of the target.Photographic device 20 can be high definition number
Code camera or other photographic devices, photographic device 20 can be set in any suitable position for being conducive to shooting, in some embodiments
In, photographic device 20 is installed on the bottom of fuselage 10 by holder.
Tracking chip 30 can track target according to clarification of objective, wherein in some embodiments, the mesh
Target feature can be the target frame that frame choosing is carried out to target.For example, person of low position indicates target, the person of low position that dotted line frame frames in Fig. 3
Indicate target frame.It further include electronic equipment 300 in the application scenes of unmanned plane 100, target frame can be set by electronics
Standby 300 are sent to unmanned plane 100.Specifically, electronic equipment 300 can show the picture that unmanned plane 100 is shot, by user to figure
Target in piece carries out frame choosing, obtains initial target frame, the initial target frame is then uploaded to unmanned plane 100.
Wherein, electronic equipment 300 is such as smart phone, tablet computer, computer, remote controler.User can be by any
Suitable type, one or more kinds of user interaction devices are interacted with electronic equipment 300, these user interaction devices can be mouse
Mark, key, touch screen etc..It, can be by being separately positioned on the wireless of respective inside between unmanned plane 100 and electronic equipment 300
Communication module (such as signal receiver, sender unit etc.) establishes communication connection, uploads or issue data/commands.Another
In some embodiments, initial target frame also be can be stored in advance in the storage device or tracking chip 30 of unmanned plane 100.
In wherein some embodiments, it can be based on the initial target frame training trace model, obtain tracker, tracking
Chip 30 utilizes the target in image to be detected of the tracker detection acquisition of photographic device 20.Some situations wherein, to
It include the target in detection image, tracking chip 30 can determine that position of the target in described image to be detected, thus
Obtain the actual position of the target.And in other situations, target be blocked or target distortion etc. due to will lead to
It can't detect the target in described image to be detected, will be unable to the actual position for obtaining the target in this case.
In order to avoid such case generation, the target detection model based on deep learning being obtained ahead of time can use to institute
It states image to be detected to be identified, obtains at least one candidate target frame and the candidate target frame pair in described image to be detected
The classification answered.Then candidate target frame is selected from least one candidate target frame according to the color characteristic of target and classification, weight
New training trace model, updates the tracker.The embodiment of the present invention updates tracker using the candidate target frame regained,
The detectability of tracker can be improved, reduce BREAK TRACK rate.
In other embodiments of unmanned plane 100, it can also be not provided with individually tracking chip 30, tracking chip 30,
Performed method (can be please referred to by one or more other controller to execute in flight control system and path planning system
Controller 40 in Fig. 4).It is executed by one or more other controller: the mapping to be checked shot based on photographic device 20
Picture determines position of the target in image to be detected, to obtain the actual position of the target, and based on tracking mesh
Target position plans the flight path of unmanned plane, and control unmanned plane flight attitude so that unmanned plane by specified path
Flight etc..
Fig. 5 is a kind of flow diagram of method for tracking target provided in an embodiment of the present invention, and the method can be by Fig. 1
Middle unmanned plane 100 execute (specifically, in some embodiments, the method by tracking system in unmanned plane 100 tracking chip
30 execute, and in further embodiments, the method is executed by the controller 40 of unmanned plane 100), as shown in figure 5, the method
Include:
101: obtaining image to be detected.
Wherein, described image to be detected is obtained by the photographic device 20 of unmanned plane 100, is shot in photographic device 20 described
Before image to be detected, unmanned plane 100 need to be by photographic device 20 towards the direction where target.
102: detecting the target in described image to be detected using tracker, wherein the tracker is based on the target
Feature training obtain.
Wherein, in some embodiments, tracker can be direct after other devices are obtained by training trace model
Load is on unmanned plane 100.In further embodiments, tracker is that unmanned plane 100 is obtained self by training trace model
, in this embodiment, before detecting the target in described image to be detected using tracker, the method also includes instructions
The step of practicing tracker, it may be assumed that obtain initial target frame, be based on the initial target frame training trace model, obtain tracker.Its
In, initial target frame, which can be, to be previously stored on unmanned plane 100, is also possible to electronic equipment 300 and is uploaded to unmanned plane 100
's.
Wherein, in some embodiments, the tracker is based on core correlation filtering (Kernel Correlation
Filter, KCF) tracker.In other embodiments, other correlation filtering trackers can also be used.It is tracked below with KCF
Illustrate tracker training for device and detects the principle of target in described image to be detected using tracker.
Please refer to Fig. 6 a, first to the initial target frame (frame that dotted line indicates is positive sample) carry out cyclic shift it
After obtain multiple sample panes, the corresponding label of sample pane is according to the far and near assignment apart from positive sample, closer, the label value of distance
It is bigger.By multiple sample panes and its corresponding label training trace model, tracker is obtained.Fig. 6 b is please referred to, tracking is being utilized
When device detects the target in image to be detected, cyclic shift first is carried out to dotted line frame, obtains multiple frames to be detected.Then
The response of each frame to be detected is calculated using tracker, the position of the maximum frame to be detected of response is target to be detected
Position in image, it is possible thereby to obtain the actual position of target.In figure 6b, the maximum frame to be detected of response should be overstriking
The frame that line indicates.
103: judging the target whether is detected in described image to be detected;
It is blocked or the situations such as target distortion in target, the target will be can't detect in described image to be detected.?
In some of embodiments (such as embodiment using KCF tracker detection target), institute can be detected according to using tracker
The maximum response for stating image to be detected acquisition judges whether to detect the target in image to be detected.If peak response
Be worth larger, be greater than default response lag, then can determine and detect the target in described image to be detected, and will it is described most
Big position of the corresponding position of response as the target in described image to be detected.If maximum response be less than or
Equal to default response lag, it is determined that do not detect the target in described image to be detected.Wherein it is possible to by multiple
Test, selection target detection effect are preferably worth as default response lag.
104: if the target is not detected in described image to be detected, the input of described image to be detected being based on
The goal-selling detection model of deep learning, to obtain at least one candidate target frame and the corresponding class of the candidate target frame
Not.
If not detecting target in image to be detected using tracker, image to be detected is inputted into goal-selling
Detection model is identified that the multiple candidate target frames and candidate target frame for obtaining each target in image to be detected are corresponding
Classification.
Wherein, goal-selling detection model can be other devices by training the neural network model based on deep learning
It obtains and is loaded directly on unmanned plane 100.In further embodiments, goal-selling detection model be unmanned plane 100 from
Neural network model of the body by training based on deep learning obtains.Goal-selling detection model can pass through great amount of samples number
Accordingly and the corresponding label of sample data (i.e. classification) training obtains, such as based on the data training on data set PASCAL VOC
It obtains.In wherein some embodiments, goal-selling detection model is based on SSD (Single Shot MultiBox
Detector) the network model of algorithm.In further embodiments, it can also be replaced by other deep learning networks, for example,
YOLO (You Only Look Once), Fast-RCNN (Regions with CNN) etc..
105: selecting candidate mesh from least one described candidate target frame according to the classification of the target and color characteristic
Mark frame.
For the same target, classification and color be it is constant, can choose according to target category and color characteristic
Candidate target frame corresponding with the target out.Wherein, in some embodiments, first from least one described candidate target frame
Select candidate target frame identical with the target category.Face is selected from candidate target frame identical with the target category again
The color characteristic similarity of color characteristic and the target is maximum and is greater than a candidate target frame of default similarity threshold.Use
The color characteristic of the color characteristic matching target of candidate target frame, presets if the similarity of most like candidate target frame is greater than
Similarity threshold then updates the tracker using the candidate target frame.Wherein it is possible to by repeatedly testing, selection target phase
Preferably it is worth like effect as default similarity threshold.
Specifically, the Euclidean distance between the color characteristic of target and the color characteristic of candidate target frame, institute can be calculated
It is smaller to state Euclidean distance, then candidate target frame is more matched with the target, i.e., candidate target frame is more similar to the target.It is described
Euclidean distance is bigger, then candidate target frame is more mismatched with the target, i.e. candidate target frame and the target is more dissimilar.
Wherein, in some embodiments, the color characteristic includes Color Statistical histogram.Obtaining initial target frame
When, the Color Statistical histogram that the initial target frame obtains target can be directly based upon.The Color Statistical histogram can be with
It is the Color Statistical histogram of tri- channel middles of R, G, B point or whole channels.It, can be by color value for each channel
0-255 is quantified by step-length, such as is quantified as 0-31 by step-length 8.Then the corresponding image of initial target frame is cut,
It is divided into m multiplied by n fritter, counts the number for belonging to each color value in each fritter, that is, obtain the Color Statistical of the target
Histogram.Likewise, obtain candidate target frame Color Statistical histogram, the Color Statistical histogram of matching candidate target frame and
The Color Statistical histogram of the target, it can obtain the similarity of the two.
106: based on the candidate target frame re -training trace model chosen, updating the tracker.
The embodiment of the present invention first passes through the tracker being obtained ahead of time and detects target in image to be detected, if not described
The target is detected in image to be detected, then described image to be detected is detected by goal-selling detection model, obtained
Obtain at least one candidate target frame and the corresponding classification of the candidate target frame in image to be detected.Then according to the target
Color characteristic and classification select a candidate target frame from least one described candidate target frame, re -training tracks mould
Type obtains new tracker.I.e. when that can not detect the target in image to be detected using tracker, target detection mould is utilized
The candidate target frame that type obtains updates the tracker.The detectability of tracker can be improved, reduce BREAK TRACK rate.
In practical applications, described image to be detected is sequential frame image, and the detection to image to be detected is to continue progress
's.For any frame image to be detected, the target in image to be detected is detected first with tracker, if be not detected described
Target then obtains candidate target frame using goal-selling detection model, and updates the tracker based on the candidate target frame,
Recycle updated tracker detection next frame image to be detected.If candidate has not been obtained using goal-selling detection model
Target frame then continues to detect next frame image to be detected with goal-selling detection model.It is urged with such, if goal-selling detects
Model there is no to candidate target frame for a long time, then initial target frame, and base need to be reselected on an electronic device by user
Tracker is updated in the initial target frame.
Specifically, in some embodiments, when can be pre-designed, and initial value be assigned, if examined from using goal-selling
And color characteristic similarity identical as the target category is not obtained at least one candidate target frame that survey model obtains to be greater than
The candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtain new image to be detected and be based on presetting
Target detection model is detected again.If obtained from least one described candidate target frame and the target category phase
With and color characteristic maximum similarity be greater than the candidate target frame of default similarity threshold, then reset when will be pre-designed.When described
Numerical value when pre-designed reaches preset threshold, then reacquires initial target frame, and based on described in initial target frame update
Tracker.Wherein, clocking capability can be realized by counter or timer.Preset threshold can take any suitable number
Value, such as it is equivalent to the numerical value of 30 seconds or 1 minute.
Correspondingly, as shown in fig. 7, described device can be used the embodiment of the invention also provides a kind of target tracker
In unmanned plane shown in FIG. 1, target tracker 700 includes:
Image collection module 701, for obtaining image to be detected;
Tracker detection module 702, for detecting the target in described image to be detected using tracker, wherein described
Tracker is based on clarification of objective training and obtains;
Judgment module 703, for judging whether detect the target in described image to be detected;
Module of target detection 704, if for being not detected the target in described image to be detected, will it is described to
Detection image inputs the goal-selling detection model based on deep learning, to obtain at least one candidate target frame and the time
Select the corresponding classification of target frame;
Candidate target frame selecting module 705, for according to the classification of the target and color characteristic from it is described at least one
A candidate target frame is selected in candidate target frame;
First tracker update module 706, for based on the candidate target frame re -training trace model chosen, to update
The tracker.
The embodiment of the present invention first passes through the tracker being obtained ahead of time and detects target in image to be detected, if not described
The target is detected in image to be detected, then described image to be detected is detected by goal-selling detection model, obtained
Obtain at least one candidate target frame and the corresponding classification of the candidate target frame in image to be detected.Then according to the target
Color characteristic and classification select a candidate target frame from least one described candidate target frame, re -training tracks mould
Type obtains new tracker.I.e. when that can not detect the target in image to be detected using tracker, target detection mould is utilized
The candidate target frame that type obtains updates the tracker.The detectability of tracker can be improved, reduce BREAK TRACK rate.
In wherein some embodiments, judgment module 703 is specifically used for:
Judge to detect whether the maximum response that described image to be detected obtains is not more than default sound using the tracker
Answer threshold value;
If so, the target is not detected in determination in described image to be detected.
In wherein some embodiments, candidate target frame selecting module 705 is specifically used for:
Candidate target frame identical with the target category is selected from least one described candidate target frame;
The color characteristic phase of color characteristic and the target is selected from candidate target frame identical with the target category
Candidate target frame that is maximum like degree and being greater than default similarity threshold.
In wherein some embodiments, judgment module 703 also particularly useful for:
Judge to detect whether the maximum response that described image to be detected obtains is greater than default response using the tracker
Threshold value;
If so, determination detects the target in described image to be detected.
In wherein some embodiments, Fig. 8, described device are please referred to further include:
Target position determining module 710 is used for using the corresponding position of the maximum response as the target described
Position in image to be detected.
In wherein some embodiments, Fig. 8 is please referred to, the clarification of objective includes the initial target frame of the target;
Described device further includes tracker training module 707, for detected using tracker the target in described image to be detected it
Before:
Obtain initial target frame;
Based on the initial target frame training trace model, tracker is obtained.
In wherein some embodiments, the color characteristic includes Color Statistical histogram.
In wherein some embodiments, Fig. 8, described device are please referred to further include:
Color of object feature obtains module 708, for obtaining the Color Statistical of the target based on the initial target frame
Histogram.
In wherein some embodiments, Fig. 8 is please referred to, described device further includes the second tracker update module 709, is used
In:
If it is similar not obtain and color characteristic identical as the target category from least one described candidate target frame
Degree is greater than the candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtains new image to be detected base
It is detected again in goal-selling detection model;
If obtaining and color characteristic maximum phase identical as the target category from least one described candidate target frame
It is greater than the candidate target frame of default similarity threshold like degree, then resets when will be pre-designed;
If numerical value when described pre-designed reaches preset threshold, initial target frame is reacquired, and based on described first
Beginning target frame updates the tracker.
In wherein some embodiments, the tracker is the tracker based on core correlation filtering.
In wherein some embodiments, the goal-selling detection model is the target detection model based on SSD algorithm.
Method for tracking target described in any of the above-described embodiment can by unmanned plane 100 tracking chip 30 or control
Device 40 executes, and tracking chip 30 (referring to figure 2.) or controller 40 (referring to figure 4.) can use hardware knot as shown in Figure 9
Structure.As shown in figure 9, the hardware configuration includes:
One or more processors 1 and memory 2, in Fig. 9 by taking a processor 1 as an example.
Processor 1 can be connected with memory 2 by bus or other modes, in Fig. 9 for being connected by bus.
Memory 2 be used as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software program,
Non-volatile computer executable program and module, as the corresponding program of method for tracking target in the embodiment of the present application refers to
Order/module is (for example, attached image collection module shown in Fig. 7 701, tracker detection module 702, judgment module 703, target are examined
Survey module 704, candidate target frame selecting module 705 and the first tracker update module 706).Processor 1 is stored in by operation
Non-volatile software program, instruction and module in memory 2 are answered thereby executing controller or the various functions of tracking chip
With and data processing, that is, realize above method embodiment method for tracking target.
Memory 2 may include storing program area and storage data area, wherein storing program area can storage program area,
Application program required at least one function;Storage data area, which can be stored, uses created data etc. according to controller.
It can also include nonvolatile memory in addition, memory 2 may include high-speed random access memory, for example, at least one
Disk memory, flush memory device or other non-volatile solid state memory parts.In some embodiments, the optional packet of memory 2
The memory remotely located relative to processor 1 is included, these remote memories can pass through network connection to unmanned plane.Above-mentioned net
The example of network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
One or more of modules are stored in the memory 2, are held when by one or more of processors 1
When row, the method for tracking target in above-mentioned any means embodiment is executed, for example, executing the method step in Fig. 5 described above
Rapid 101 to step 106;Realize the function of the module 701-710 in module 701-706, Fig. 8 in Fig. 7.
Method provided by the embodiment of the present application can be performed in the said goods, has the corresponding functional module of execution method and has
Beneficial effect.The not technical detail of detailed description in the present embodiment, reference can be made to method provided by the embodiment of the present application.
The embodiment of the present application provides a kind of non-volatile computer readable storage medium storing program for executing, the computer-readable storage medium
Matter is stored with computer executable instructions, which is executed by one or more processors, such as in Fig. 9
One processor 1 may make said one or multiple processors that the target following side in above-mentioned any means embodiment can be performed
Method, for example, executing method and step 101 in Fig. 5 described above to step 106;Realize module 701-706, Fig. 8 in Fig. 7
In module 701-710 function.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.
By the description of above embodiment, those of ordinary skill in the art can be understood that each embodiment can borrow
Help software that the mode of general hardware platform is added to realize, naturally it is also possible to pass through hardware.Those of ordinary skill in the art can manage
Solution realize above-described embodiment method in all or part of the process be can be instructed by computer program relevant hardware come
It completes, the program can be stored in a computer-readable storage medium, and the program is when being executed, it may include such as above-mentioned each
The process of the embodiment of method.Wherein, the storage medium can be magnetic disk, CD, read-only memory (Read-Only
Memory, ROM) or random access memory (RandomAccessMemory, RAM) etc..
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;At this
It under the thinking of invention, can also be combined between the technical characteristic in above embodiments or different embodiment, step can be with
It is realized with random order, and there are many other variations of different aspect present invention as described above, for simplicity, they do not have
Have and is provided in details;Although the present invention is described in detail referring to the foregoing embodiments, the ordinary skill people of this field
Member is it is understood that it is still possible to modify the technical solutions described in the foregoing embodiments, or to part of skill
Art feature is equivalently replaced;And these are modified or replaceed, each reality of the present invention that it does not separate the essence of the corresponding technical solution
Apply the range of a technical solution.
Claims (24)
1. a kind of method for tracking target, which is characterized in that the described method includes:
Obtain image to be detected;
The target in described image to be detected is detected using tracker, wherein the tracker is instructed based on the clarification of objective
Practice and obtains;
Judge the target whether is detected in described image to be detected;
If the target is not detected in described image to be detected, the input of described image to be detected is based on deep learning
Goal-selling detection model, to obtain at least one candidate target frame and the corresponding classification of the candidate target frame;
Candidate target frame is selected from least one described candidate target frame according to the classification of the target and color characteristic;
Based on the candidate target frame re -training trace model chosen, the tracker is updated.
2. the method according to claim 1, wherein whether the judgement detects in described image to be detected
The target, comprising:
Judge to detect whether the maximum response that described image to be detected obtains is not more than default threshold of response using the tracker
Value;
If so, the target is not detected in determination in described image to be detected.
3. method according to claim 1 or 2, which is characterized in that the classification and color characteristic according to the target
A candidate target frame is selected from least one described candidate target frame, comprising:
Candidate target frame identical with the target category is selected from least one described candidate target frame;
The color characteristic similarity of color characteristic and the target is selected from candidate target frame identical with the target category
Candidate target frame that is maximum and being greater than default similarity threshold.
4. the method according to claim 1, wherein whether the judgement detects in the feature detection image
To the target, comprising:
Judge to detect whether the maximum response that described image to be detected obtains is greater than default response lag using the tracker;
If so, determination detects the target in described image to be detected.
5. according to the method described in claim 4, it is characterized in that, this method further include:
Position using the corresponding position of the maximum response as the target in described image to be detected.
6. method according to any one of claims 1-5, which is characterized in that the clarification of objective includes the target
Initial target frame, then, before the target detected using tracker in described image to be detected, the method is also wrapped
It includes:
Obtain initial target frame;
Based on the initial target frame training trace model, the tracker is obtained.
7. according to the method described in claim 3, it is characterized in that, the color characteristic includes Color Statistical histogram.
8. according to the method described in claim 6, it is characterized in that, the method also includes:
The Color Statistical histogram of the target is obtained based on the initial target frame.
9. according to the method described in claim 3, it is characterized in that, the method also includes:
If it is big not obtain and color characteristic similarity identical as the target category from least one described candidate target frame
In the candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtain new image to be detected be based on it is pre-
If target detection model is detected again;
If obtaining and color characteristic maximum similarity identical as the target category from least one described candidate target frame
It is reset greater than the candidate target frame of default similarity threshold, then when will be pre-designed;
If numerical value when described pre-designed reaches preset threshold, initial target frame is reacquired, and be based on the initial mesh
It marks frame and updates the tracker.
10. method according to claim 1 to 9, which is characterized in that the tracker is to close to filter based on nuclear phase
The tracker of wave algorithm.
11. -10 described in any item methods according to claim 1, which is characterized in that the goal-selling detection model be based on
The target detection model of SSD algorithm.
12. a kind of target tracker, which is characterized in that described device includes:
Image collection module, for obtaining image to be detected;
Tracker detection module, for detecting the target in described image to be detected using tracker, wherein the tracker base
It is obtained in clarification of objective training;
Judgment module, for judging whether detect the target in described image to be detected;
Module of target detection, if for the target to be not detected in described image to be detected, by the mapping to be checked
As goal-selling detection model of the input based on deep learning, to obtain at least one candidate target frame and the candidate target
The corresponding classification of frame;
Candidate target frame selecting module, for according to the classification of the target and color characteristic from least one described candidate target
Candidate target frame is selected in frame;
First tracker update module, for updating the tracking based on the candidate target frame re -training trace model chosen
Device.
13. device according to claim 12, which is characterized in that the judgment module is specifically used for:
Judge to detect whether the maximum response that described image to be detected obtains is not more than default threshold of response using the tracker
Value;
If so, the target is not detected in determination in described image to be detected.
14. device according to claim 12 or 13, which is characterized in that the candidate target frame selecting module is specifically used
In:
Candidate target frame identical with the target category is selected from least one described candidate target frame;
The color characteristic similarity of color characteristic and the target is selected from candidate target frame identical with the target category
Candidate target frame that is maximum and being greater than default similarity threshold.
15. device according to claim 12, which is characterized in that the judgment module also particularly useful for:
Judge to detect whether the maximum response that described image to be detected obtains is greater than default response lag using the tracker;
If so, determination detects the target in described image to be detected.
16. device according to claim 15, which is characterized in that described device further include:
Target position determining module is used for using the corresponding position of the maximum response as the target in the mapping to be checked
Position as in.
17. device described in any one of 2-16 according to claim 1, which is characterized in that the clarification of objective includes described
The initial target frame of target;
Described device further includes tracker training module, for detecting the mesh in described image to be detected using the tracker
Before mark:
Obtain initial target frame;
Based on the initial target frame training trace model, the tracker is obtained.
18. device according to claim 14, which is characterized in that the color characteristic includes Color Statistical histogram.
19. device according to claim 17, which is characterized in that described device further include:
Color of object feature obtains module, for obtaining the Color Statistical histogram of the target based on the initial target frame.
20. device according to claim 14, which is characterized in that described device further includes the second tracker update module,
For:
If it is big not obtain and color characteristic similarity identical as the target category from least one described candidate target frame
In the candidate target frame of default similarity threshold, then numerical value when increasing pre-designed, and obtain new image to be detected be based on it is pre-
If target detection model is detected again;
If obtaining and color characteristic maximum similarity identical as the target category from least one described candidate target frame
It is reset greater than the candidate target frame of default similarity threshold, then when will be pre-designed;
If numerical value when described pre-designed reaches preset threshold, initial target frame is reacquired, and be based on the initial mesh
It marks frame and updates the tracker.
21. device described in any one of 2-20 according to claim 1, which is characterized in that the tracker is to be closed based on nuclear phase
The tracker of filtering algorithm.
22. device described in any one of 2-21 according to claim 1, which is characterized in that the goal-selling detection model is
Target detection model based on SSD algorithm.
23. a kind of unmanned plane, which is characterized in that the unmanned plane includes fuselage, the horn being connected with the fuselage, set on described
The dynamical system of horn, the photographic device for being set to the fuselage and tracking chip, the photographic device and the tracking chip
It is electrically connected, wherein for obtaining image to be detected, the tracking chip includes: the photographic device
At least one processor;And
The memory being connect at least one described processor communication;Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
It manages device to execute, so that at least one described processor is able to carry out the described in any item methods of claim 1-11.
24. a kind of non-volatile computer readable storage medium storing program for executing, which is characterized in that the computer-readable recording medium storage has
Computer executable instructions make the unmanned plane execute such as right when the computer executable instructions are executed by unmanned plane
It is required that the described in any item methods of 1-11.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910213970.2A CN109978045A (en) | 2019-03-20 | 2019-03-20 | A kind of method for tracking target, device and unmanned plane |
PCT/CN2020/078629 WO2020187095A1 (en) | 2019-03-20 | 2020-03-10 | Target tracking method and apparatus, and unmanned aerial vehicle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910213970.2A CN109978045A (en) | 2019-03-20 | 2019-03-20 | A kind of method for tracking target, device and unmanned plane |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109978045A true CN109978045A (en) | 2019-07-05 |
Family
ID=67079721
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910213970.2A Pending CN109978045A (en) | 2019-03-20 | 2019-03-20 | A kind of method for tracking target, device and unmanned plane |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN109978045A (en) |
WO (1) | WO2020187095A1 (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110569840A (en) * | 2019-08-13 | 2019-12-13 | 浙江大华技术股份有限公司 | Target detection method and related device |
CN110687922A (en) * | 2019-11-08 | 2020-01-14 | 湖北经济学院 | Visual tracking method of unmanned aerial vehicle and unmanned aerial vehicle with visual tracking function |
CN111242981A (en) * | 2020-01-21 | 2020-06-05 | 北京捷通华声科技股份有限公司 | Tracking method and device for fixed object and security equipment |
CN111667505A (en) * | 2020-04-30 | 2020-09-15 | 北京捷通华声科技股份有限公司 | Method and device for tracking fixed object |
WO2020187095A1 (en) * | 2019-03-20 | 2020-09-24 | 深圳市道通智能航空技术有限公司 | Target tracking method and apparatus, and unmanned aerial vehicle |
CN113256680A (en) * | 2021-05-13 | 2021-08-13 | 燕山大学 | High-precision target tracking system based on unsupervised learning |
CN113470078A (en) * | 2021-07-15 | 2021-10-01 | 浙江大华技术股份有限公司 | Target tracking method, device and system |
CN114937231A (en) * | 2022-07-21 | 2022-08-23 | 成都西物信安智能系统有限公司 | Method for improving target identification tracking accuracy |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112700469A (en) * | 2020-12-30 | 2021-04-23 | 武汉卓目科技有限公司 | Visual target tracking method and device based on ECO algorithm and target detection |
CN112686921B (en) * | 2021-01-08 | 2023-12-01 | 西安羚控电子科技有限公司 | Multi-interference unmanned aerial vehicle detection tracking method based on track characteristics |
CN112733741A (en) * | 2021-01-14 | 2021-04-30 | 苏州挚途科技有限公司 | Traffic signboard identification method and device and electronic equipment |
CN113808170B (en) * | 2021-09-24 | 2023-06-27 | 电子科技大学长三角研究院(湖州) | Anti-unmanned aerial vehicle tracking method based on deep learning |
CN113807464B (en) * | 2021-09-29 | 2022-05-13 | 东南大学 | Unmanned aerial vehicle aerial image target detection method based on improved YOLO V5 |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106960446A (en) * | 2017-04-01 | 2017-07-18 | 广东华中科技大学工业技术研究院 | A kind of waterborne target detecting and tracking integral method applied towards unmanned boat |
CN107564034A (en) * | 2017-07-27 | 2018-01-09 | 华南理工大学 | The pedestrian detection and tracking of multiple target in a kind of monitor video |
CN107563313A (en) * | 2017-08-18 | 2018-01-09 | 北京航空航天大学 | Multiple target pedestrian detection and tracking based on deep learning |
CN107918765A (en) * | 2017-11-17 | 2018-04-17 | 中国矿业大学 | A kind of Moving target detection and tracing system and its method |
CN107943837A (en) * | 2017-10-27 | 2018-04-20 | 江苏理工学院 | A kind of video abstraction generating method of foreground target key frame |
CN108229442A (en) * | 2018-02-07 | 2018-06-29 | 西南科技大学 | Face fast and stable detection method in image sequence based on MS-KCF |
CN108596955A (en) * | 2018-04-25 | 2018-09-28 | Oppo广东移动通信有限公司 | A kind of image detecting method, image detection device and mobile terminal |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9443137B2 (en) * | 2012-05-08 | 2016-09-13 | Samsung Electronics Co., Ltd. | Apparatus and method for detecting body parts |
CN107066990B (en) * | 2017-05-04 | 2019-10-11 | 厦门美图之家科技有限公司 | A kind of method for tracking target and mobile device |
CN109409354B (en) * | 2017-08-18 | 2021-09-21 | 深圳市道通智能航空技术股份有限公司 | Unmanned aerial vehicle intelligent following target determination method, unmanned aerial vehicle and remote controller |
CN108062764A (en) * | 2017-11-30 | 2018-05-22 | 极翼机器人(上海)有限公司 | A kind of object tracking methods of view-based access control model |
CN109978045A (en) * | 2019-03-20 | 2019-07-05 | 深圳市道通智能航空技术有限公司 | A kind of method for tracking target, device and unmanned plane |
-
2019
- 2019-03-20 CN CN201910213970.2A patent/CN109978045A/en active Pending
-
2020
- 2020-03-10 WO PCT/CN2020/078629 patent/WO2020187095A1/en active Application Filing
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106960446A (en) * | 2017-04-01 | 2017-07-18 | 广东华中科技大学工业技术研究院 | A kind of waterborne target detecting and tracking integral method applied towards unmanned boat |
CN107564034A (en) * | 2017-07-27 | 2018-01-09 | 华南理工大学 | The pedestrian detection and tracking of multiple target in a kind of monitor video |
CN107563313A (en) * | 2017-08-18 | 2018-01-09 | 北京航空航天大学 | Multiple target pedestrian detection and tracking based on deep learning |
CN107943837A (en) * | 2017-10-27 | 2018-04-20 | 江苏理工学院 | A kind of video abstraction generating method of foreground target key frame |
CN107918765A (en) * | 2017-11-17 | 2018-04-17 | 中国矿业大学 | A kind of Moving target detection and tracing system and its method |
CN108229442A (en) * | 2018-02-07 | 2018-06-29 | 西南科技大学 | Face fast and stable detection method in image sequence based on MS-KCF |
CN108596955A (en) * | 2018-04-25 | 2018-09-28 | Oppo广东移动通信有限公司 | A kind of image detecting method, image detection device and mobile terminal |
Non-Patent Citations (1)
Title |
---|
熊有伦: "《机器人学 建模、控制与视觉》", 31 March 2018, 华中科技大学出版社 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020187095A1 (en) * | 2019-03-20 | 2020-09-24 | 深圳市道通智能航空技术有限公司 | Target tracking method and apparatus, and unmanned aerial vehicle |
CN110569840A (en) * | 2019-08-13 | 2019-12-13 | 浙江大华技术股份有限公司 | Target detection method and related device |
CN110687922A (en) * | 2019-11-08 | 2020-01-14 | 湖北经济学院 | Visual tracking method of unmanned aerial vehicle and unmanned aerial vehicle with visual tracking function |
CN111242981A (en) * | 2020-01-21 | 2020-06-05 | 北京捷通华声科技股份有限公司 | Tracking method and device for fixed object and security equipment |
CN111667505A (en) * | 2020-04-30 | 2020-09-15 | 北京捷通华声科技股份有限公司 | Method and device for tracking fixed object |
CN111667505B (en) * | 2020-04-30 | 2023-04-07 | 北京捷通华声科技股份有限公司 | Method and device for tracking fixed object |
CN113256680A (en) * | 2021-05-13 | 2021-08-13 | 燕山大学 | High-precision target tracking system based on unsupervised learning |
CN113470078A (en) * | 2021-07-15 | 2021-10-01 | 浙江大华技术股份有限公司 | Target tracking method, device and system |
CN114937231A (en) * | 2022-07-21 | 2022-08-23 | 成都西物信安智能系统有限公司 | Method for improving target identification tracking accuracy |
Also Published As
Publication number | Publication date |
---|---|
WO2020187095A1 (en) | 2020-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109978045A (en) | A kind of method for tracking target, device and unmanned plane | |
Rohan et al. | Convolutional neural network-based real-time object detection and tracking for parrot AR drone 2 | |
US20210271264A1 (en) | Aerial Vehicle Touchdown Detection | |
CN111932588B (en) | Tracking method of airborne unmanned aerial vehicle multi-target tracking system based on deep learning | |
US10140719B2 (en) | System and method for enhancing target tracking via detector and tracker fusion for unmanned aerial vehicles | |
CN107000832B (en) | Unmanned plane undercarriage control method, device, unmanned plane and its system | |
CN109409354B (en) | Unmanned aerial vehicle intelligent following target determination method, unmanned aerial vehicle and remote controller | |
US10671068B1 (en) | Shared sensor data across sensor processing pipelines | |
US20190325584A1 (en) | Systems and Methods for Real-Time Adjustment of Neural Networks for Autonomous Tracking and Localization of Moving Subject | |
Ludington et al. | Augmenting UAV autonomy | |
CN110909630B (en) | Abnormal game video detection method and device | |
CN110244760A (en) | A kind of barrier-avoiding method, device and electronic equipment | |
CN106292799B (en) | Unmanned plane, remote control and its control method | |
CN107871324A (en) | One kind is based on twin-channel method for tracking target and device | |
WO2021088684A1 (en) | Omnidirectional obstacle avoidance method and unmanned aerial vehicle | |
CN108121350B (en) | Method for controlling aircraft to land and related device | |
KR102330055B1 (en) | Method and system for detecting change of structure using drone | |
Maher et al. | Realtime human-UAV interaction using deep learning | |
KR102210404B1 (en) | Location information extraction device and method | |
CN106647788A (en) | Unmanned plane flight control method and device | |
CN109631853A (en) | A kind of depth map processing method, device and unmanned plane | |
KR102194127B1 (en) | Drone having MEMS sensor | |
Lu et al. | Target localization with drones using mobile CNNs | |
CN108363410A (en) | Flight control method, device and the unmanned plane of unmanned plane | |
CN107749942A (en) | Suspension image pickup method, mobile terminal and computer-readable recording medium |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 518055 Guangdong city of Shenzhen province Nanshan District Xili Street Xueyuan Road No. 1001 Chi Yuen Building 9 layer B1 Applicant after: Shenzhen daotong intelligent Aviation Technology Co.,Ltd. Address before: 518055 Guangdong city of Shenzhen province Nanshan District Xili Street Xueyuan Road No. 1001 Chi Yuen Building 9 layer B1 Applicant before: AUTEL ROBOTICS Co.,Ltd. |