CN104793200B - Dynamic planning track-before-detect method based on iterative processing - Google Patents
Dynamic planning track-before-detect method based on iterative processing Download PDFInfo
- Publication number
- CN104793200B CN104793200B CN201510164452.8A CN201510164452A CN104793200B CN 104793200 B CN104793200 B CN 104793200B CN 201510164452 A CN201510164452 A CN 201510164452A CN 104793200 B CN104793200 B CN 104793200B
- Authority
- CN
- China
- Prior art keywords
- state
- flight path
- value function
- value
- track
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/66—Radar-tracking systems; Analogous systems
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a dynamic planning track-before-detect method based on iterative processing, solves the problem that conventional dynamic planning track-before-detect algorithms based on sliding window processing are large in calculation amount, improves estimation accuracy of the algorithms to target track and lowers false alarm rate of the algorithms. The dynamic planning track-before-detect method is characterized in that when a dynamic planning algorithm is utilized to track a target, a value function of current moment is estimated according to a value function of previous moment; only two adjacent frames of data are iteratively processed at each moment, and combined processing of multiple frames of data is not needed, so that conventional sliding window processing structure is broken through, the problem that calculation amount is increased rapidly due to the fact that echo data are repeatedly processed in conventional methods is solved effectively, estimation accuracy of the algorithms to the target track is improved, value functions of noise state are reduced, realtime target tracking is guaranteed, more accurate target moving paths are provided while false alarm is inhibited, and performance of the algorithms is improved greatly.
Description
Technical field
The invention belongs to Radar Targets'Detection field, more particularly to radar is led to tracking technique before dim target detection
Domain.
Background technology
Current radar is faced with stern challenge to the Detection Techniques of target.The reflection cross section of Stealthy Target is little, target
Apart from radar farther out and the mountain region that runs in practical application, the strong clutter background such as city all causes target echo relative to clutter
Very weak, the signal to noise ratio of target is greatly reduced, it is difficult to be detected.
It is a kind of method of effective detecting and tracking weak target based on track algorithm before the detection of Dynamic Programming.It passes through
Accumulation Combined Treatment multiframe echo data to value function, is a kind of equivalent realization of exhaustive search, but its efficiency far is higher than
Exhaustive search.Recently, the method is widely used in the fields such as radar tracking, image procossing.Although it can effectively be detected
With tracking weak target, but because it is required for calculating the value function of all states at each moment, when the echo for receiving
When data volume is very big, the amount of calculation needed for the method can increase sharply.And because dynamic programming algorithm is a kind of batch processing
Method, in order to track target for a long time, needs to combine slide window processing technology in actual applications, causes every frame echo data to be weighed
It is multiple to process multiple, further increase the computation burden of algorithm, it is difficult to realize the real-time tracking to target.Particularly when the letter of target
When making an uproar relatively low frequently, in order to improve the detectability to target, the data frame number processed in a sliding window increases, the calculating of algorithm
Amount also can be sharply increased therewith, substantially prolongs the signal processing time of system, the real-time of system be reduced, to the detection of target
Track band carrys out very big difficulty.Therefore, in order to meet radar detection requirement of real-time, research can reduce the dynamic rule of amount of calculation
Method to one's profit is very crucial.Chinese scholars have also done certain research to this problem.Document " Thresholding
Process Based Dynamic Programming Track-Before-Detect Algorithm,IEICE Trans
Commun, vol.E96.B, 291-300,2013 " detects frame data using the first low threshold, and screening is probably the point of target
Mark, then processes these discrete point marks using dynamic programming algorithm, recovers targetpath；Document " A new TBD-DP
Algorithm using multiple IR sensors to locate the target launch point, proc of
SPIE, vol.8185, p81850P, 2011 " proposes a kind of resolution ratio for reducing reception image to reduce the number of pending data
The dynamic programming algorithm of amount.But these methods are only reduction of the data processing amount in single sliding window, can not inherently solve
The problem that echo data is computed repeatedly caused by multiple slide window processing is certainly performed, the calculating of algorithm in long-time tracking target
Complexity is appointed so very high, it is difficult to ensure the real-time of system.
The content of the invention
The present invention for background technology technical problem not enough to be solved be to provide a kind of amount of calculation it is little, detection more
The value function accumulation method tracked before accurate multi frame detection rapidly and efficiently.
The present invention is for the solution technical scheme that adopted of above-mentioned technical problem, a kind of Dynamic Programming based on iterative processing
Tracking before detection, including step once：
Step 1, initialization system parameter：
Initialization observation totalframes M, the sliding window length of track algorithm is K before Dynamic Programming detection, imitative using Monte Carlo
True experiment calculates detection threshold VT, initializing variable k=1；
Step 2, the reading kth frame measurement from radar receiver；
Step 3, accumulation dynamic planning value function：
If k=1, then initialize the state with the range value of the corresponding first frame echo data of each quantization state
Value function；
If k ≠ 1, then each value function for quantifying state is the value letter of the previous frame state that possible be transferred to the state
Several maximums range value sum corresponding with the state；
If step 4, k<K, then make k=k+1, return to step 2；Otherwise, execution step 5；
Step 5, Threshold detection：The value function that step 4 is obtained and thresholding VTRelatively, if being more than thresholding, then it is assumed that the shape
State is candidate target state, is deposited into set D；
Step 6, flight path recover：
To each candidate target state in set D, flight path recovery is carried out, in set T, then the flight path after recovery is stored in
Make D=Φ；
Step 7, track association：
If k=K, the new flight path of the short track initiation per bar one in set T is stored in set T*In, execution step 8；
If k>K, initializes i=1, j=0；
If 7.1. j<N, N are set T*In flight path number, then j=j+1, execution step 7.2；Otherwise, with set T
I-th flight path tiAn initial new flight path, is stored in setIn, then execution step 7.3；
If 7.2. T*In j-th strip flight pathAssociate, then, return to step 7.1；IfWithout association, will gather
I-th flight path t in TiWithIt is compared, if error of two flight paths at k-K+1～k-1 moment is given in advance less than one
Fixed empirical value, then, they are associated, use tiLast state updateAnd recordTo associate, step is then performed
Rapid 7.3；If the error of two flight paths exceeds empirical value, then, return to step 7.1；
If 7.3. i<N ', N ' are the flight path number in set T, then, i=i+1, j=0, return to step 7.1；Otherwise, navigate
Mark association terminates, willIt is incorporated to T*, Ran Houling
Step 8, output T*In flight path；
Step 9, estimated state value function：
If k<M, by the value function that step 4 is obtained the value letter of kth-K+1 frames (the first frame i.e. now in sliding window) is deducted
Number, obtains new value function；Then k=k+1, return to step 2 are made；
If k >=M, stop calculating.
If k ≠ 1 in the step 3, then the value for the previous frame state of the state may be transferred to of the value function of each state
The maximum of function range value sum corresponding with the state, i.e.,
And record xkCorresponding previous frame state
It is Z that wherein kth frame is measuredk={ zk(m,n),1≤m≤Nx,1≤n≤Ny, zk(m, n) represents kth frame echo data
Range value in measurement unit (m, n), Nx,NyThe respectively number of X-direction and Y-direction quantifying unit, Zl:k,(l<K) l is represented
The echo data set at～k moment, xkArbitrary quantization state of kth frame is represented, I () represents state xkValue function, τ (xk)
Express possibility and be transferred to xkThe moment of kth -1 state set, z (xk) represent state xkCorresponding range value.
If k in the step 9<M, each quantization state x to kth framek, make i=k, k-1 ... k-K+2 utilize ψ (xi)
The state transfer relationship of middle record finds xkState x at corresponding k-K+1 momentk-K+1；Then, x is estimatedkValue function：
I(xk|Zk-K+2:k)=I (xk|Zk-K+1:k)-z(xk-K+1), make k=k+1, return to step 2.
It is an advantage of the invention that the dynamic programming algorithm amount of calculation required in long-time tracking target is greatly reduced,
While real-time tracking target, there is provided more accurate target motion path, and suppress the generation of false-alarm, effectively improve algorithm
Performance.Present invention could apply to the field such as radar tracking, image procossing.
Description of the drawings：
Fig. 1 is the FB(flow block) of the present invention.
Fig. 2 is the amount of calculation comparison diagram of the present invention and the front track algorithm of Dynamic Programming detection based on slide window processing.
Fig. 3 is the flight path and true flight path that the present invention recovers with the front track algorithm of Dynamic Programming detection based on slide window processing
Position root-mean-square error comparison diagram.
Fig. 4 is that the present invention at a time accumulates what is obtained with the front track algorithm of Dynamic Programming detection based on slide window processing
The comparison diagram of value function；A () is the value function that the present invention is calculated, (b) be tracking before the Dynamic Programming detection based on slide window processing
The value function that algorithm is calculated, is (c) difference of (a) and (b).
Specific embodiment：
The main method using Computer Simulation of the invention is verified that all steps, conclusion are all in MATLAB-R2010b
Upper checking confirms.Specific implementation step is as follows：
Step 1, initialization system parameter：
Initialization observation totalframes M, Dynamic Programming detection before track algorithm sliding window length be K, detection threshold VT, initially
Change variable k=1.
Step 2, the reading kth frame measurement from radar receiver：
It is Z that kth frame is measuredk={ zk(m,n),1≤m≤Nx,1≤n≤Ny},zk(m, n) represents kth frame echo data in amount
The range value surveyed in unit (m, n), Nx,NyThe respectively number of X-direction and Y-direction quantifying unit.
Step 3, accumulation dynamic planning value function：
If k=1, then initialize the value letter of the state with the range value of the corresponding first frame echo data of each state
Number, i.e.,
I(x1|Z1)=z (x1)。
If k ≠ 1, then the maximum for the value function of the previous frame state of the state may be transferred to of the value function of each state
Value range value sum corresponding with the state, i.e.,
And record xkCorresponding previous frame state
Wherein, Zl:k,(l<K) the echo data set at l～k moment, x are representedkRepresent arbitrary quantization state of kth frame, I
() represents state xkValue function, τ (xk) expressing possibility is transferred to xkThe state of kth -1 set, z (xk) represent state xkIt is right
The range value answered.
If step 4, k<K, then make k=k+1, return to step 2；Otherwise, execution step 5.
Step 5, Threshold detection：
If I is (xk)≥VT, then think xkIt is candidate target state, is deposited into set D.
Step 6, flight path recover：
To each candidate state x in set Dk, flight path backtracking is carried out, to i=k, k-1 ... k-K+2 utilize ψ (xi) middle record
State transfer relationship, recovery obtain state xkCorresponding length is the flight path (x of Kk-K+1,xk-K+2,…,xk), it is stored in set T
In, then make D=Φ.
Step 7, track association：
If k=K, the new flight path of the short track initiation per bar one in set T is stored in set T*In.
If k>K, initializes i=1, j=0
If 7.1. j<N, N are set T*In flight path number, then j=j+1, execution step 7.2；Otherwise, with set T
I-th flight path tiAn initial new flight path, is stored in setIn, then execution step 7.3.
If 7.2. T*In j-th strip flight pathAssociate, then, return to step 7.1.IfWithout association, will gather
I-th flight path t in TiWithBe compared, if two flight paths k-K+1～k-1 moment error in allowed band, that
, they are associated, use tiLast state updateAnd recordTo associate, then execution step 7.3；If two
The error of bar flight path exceeds allowed band, then, return to step 7.1.
If 7.3. i<N ', N ' are the flight path number in set T, then, i=i+1, j=0, return to step 7.1；Otherwise, navigate
Mark association terminates, willIt is incorporated to T*, Ran Houling
Step 8, output T*In flight path.
Step 9, estimated state value function：
If k<M, each quantization state x to kth framek, make i=k, k-1 ... k-K+2 utilize ψ (xi) the middle shape for recording
State transfer relationship finds xkState x at corresponding k-K+1 momentk-K+1；Then, x is estimatedkValue function：
I(xk|Zk-K+2:k)=I (xk|Zk-K+1:k)-z(xk-K+1), make k=k+1, return to step 2.
If k >=M, algorithm terminates.
Fig. 2 has counted the present invention with the front track algorithm of Dynamic Programming detection based on slide window processing in data processing
Execution is compared, addition, and the number of times of multiplication, and its initiation parameter is：Observation totalframes M=80, Dynamic Programming detection before with
Sliding window length K of track algorithm takes respectively 6～15 frames, detection threshold VT=14.2465, datum plane size is Nx×Ny=180 ×
180.As shown in Fig. 2 as sliding window length increases, the amount of calculation urgency of track algorithm before the Dynamic Programming detection based on slide window processing
Increase severely and add, and the amount of calculation of the present invention substantially remains in a constant level and the dynamic rule far smaller than based on slide window processing
The amount of calculation of method to one's profit.
Fig. 3 be the present invention with based on slide window processing Dynamic Programming detection before track algorithm recovery targetpath with it is true
The position root-mean-square error comparison diagram of flight path, its initiation parameter is observation totalframes M=20, and tracking before Dynamic Programming detection is calculated
The sliding window length K=8 frame of method, detection threshold VT=18.0832, datum plane size is Nx×Ny=60 × 60.As shown in figure 3,
When target signal to noise ratio than it is relatively low when, the present invention recover targetpath precision will apparently higher than based on slide window processing Dynamic Programming
The flight path precision of algorithm.
Fig. 4 is a certain moment, and the value function that the present invention is calculated, the Dynamic Programming based on slide window processing detects front track algorithm
The value function of calculating and the difference of the two, its initiation parameter is observation totalframes M=20, and tracking before Dynamic Programming detection is calculated
The sliding window length K=6 frame of method, detection threshold VT=14.2465, datum plane size is Nx×Ny=60 × 60.As shown in figure 4,
The value function that two methods are calculated all is in " mountain peak " shape, highlights the feature of dbjective state, but their value function difference is less than
Equal to 0, illustrate to compare and conventional method, the present invention reduces the value function of noise states while target signal to noise ratio is increased,
The generation of false-alarm can effectively be suppressed.
Claims (3)
1. tracking before a kind of Dynamic Programming based on iterative processing is detected, including step once：
Step 1, initialization system parameter：
Initialization observation totalframes M, the sliding window length of track algorithm is K before Dynamic Programming detection, using Monte Carlo simulation reality
Test and calculate detection threshold VT, initializing variable k=1；
Step 2, the reading kth frame measurement from radar receiver；
Step 3, accumulation dynamic planning value function：
If k=1, then initialize the value letter of the state with the range value of the corresponding first frame echo data of each quantization state
Number；
If k ≠ 1, then each value function for quantifying state is the value function of the previous frame state that possible be transferred to the state
Maximum range value sum corresponding with the state；
If step 4, k ＜ K, then make k=k+1, return to step 2；Otherwise, execution step 5；
Step 5, Threshold detection：The value function that step 4 is obtained and thresholding VTRelatively, if being more than thresholding, then it is assumed that the state is time
Dbjective state is selected, set D is deposited into；
Step 6, flight path recover：
To each candidate target state in set D, flight path recovery is carried out, the flight path after recovery is stored in set T, then make D
=Φ；
Step 7, track association：
If k=K, the new flight path of the short track initiation per bar one in set T is stored in set T*In, execution step 8；
If k is ＞ K, i=1, j=0 are initialized；
If 7.1. j ＜ N, N is set T*In flight path number, then j=j+1, execution step 7.2；Otherwise, with set T
I bar flight path tiAn initial new flight path, is stored in setIn, then execution step 7.3；
If 7.2. T*In j-th strip flight pathAssociate, then, return to step 7.1；IfWithout association, by set T
I-th flight path tiWithIt is compared, if error of two flight paths at k-K+1～k-1 moment is less than a previously given Jing
Test value, then, they are associated, use tiLast state updateAnd recordTo associate, then execution step 7.3；
If the error of two flight paths exceeds empirical value, then, return to step 7.1；
If 7.3. i ＜ N ', N ' is the flight path number in set T, then, i=i+1, j=0, return to step 7.1；Otherwise, flight path
Association terminates, willIt is incorporated to T*, Ran HoulingT=Φ；
Step 8, output T*In flight path；
Step 9, estimated state value function：
If k is ＜ M, the value function that step 4 is obtained is deducted into the value function of kth-K+1 frames (the first frame i.e. now in sliding window),
Obtain new value function；Then k=k+1, return to step 2 are made；
If k >=M, stop calculating.
2. tracking before a kind of Dynamic Programming based on iterative processing as claimed in claim 1 is detected, it is characterised in that the step
If k ≠ 1 in rapid 3, then the maximum for the value function of the previous frame state of the state may be transferred to of the value function of each state
Value range value sum corresponding with the state, i.e.,
And record xkCorresponding previous frame state
It is Z that wherein kth frame is measuredk={ zk(m,n),1≤m≤Nx,1≤n≤Ny, zk(m, n) represents kth frame echo data in amount
The range value surveyed in unit (m, n), Nx,NyThe respectively number of X-direction and Y-direction quantifying unit, Zl:kRepresent l～k moment
Echo data set, wherein l ＜ k, xkArbitrary quantization state of kth frame is represented, I () represents state xkValue function, τ (xk)
Express possibility and be transferred to xkThe moment of kth -1 state set, z (xk) represent state xkCorresponding range value.
3. tracking before a kind of Dynamic Programming based on iterative processing as claimed in claim 1 is detected, it is characterised in that institute
State in step 9 if k is ＜ M, each quantization state x to kth framek, make i=k, k-1 ... k-K+2 utilize ψ (xi) middle record
State transfer relationship finds xkState x at corresponding k-K+1 momentk-K+1；Then, x is estimatedkValue function：I(xk|Zk-K+2:k)=
I(xk|Zk-K+1:k)-z(xk-K+1), make k=k+1, return to step 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510164452.8A CN104793200B (en) | 2015-04-09 | 2015-04-09 | Dynamic planning track-before-detect method based on iterative processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510164452.8A CN104793200B (en) | 2015-04-09 | 2015-04-09 | Dynamic planning track-before-detect method based on iterative processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104793200A CN104793200A (en) | 2015-07-22 |
| CN104793200B true CN104793200B (en) | 2017-05-10 |
Family
ID=53558164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510164452.8A Active CN104793200B (en) | 2015-04-09 | 2015-04-09 | Dynamic planning track-before-detect method based on iterative processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104793200B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105467374B (en) * | 2016-01-08 | 2018-03-20 | 西安电子科技大学 | Object detection method based on prewhitening ratio cell average detector under sea clutter background |
| CN106199584B (en) * | 2016-07-05 | 2018-12-18 | 电子科技大学 | A kind of track initiation method based on measurement storage |
| CN107340517B (en) * | 2017-07-04 | 2021-02-05 | 电子科技大学 | Multi-sensor multi-frame tracking-before-detection method |
| CN107544066B (en) * | 2017-07-09 | 2020-05-12 | 电子科技大学 | Distributed asynchronous iterative filtering fusion method based on pre-detection tracking |
| CN110376579B (en) * | 2019-07-22 | 2023-04-18 | 西安电子工程研究所 | Dynamic programming track-before-detect method for maneuvering target |
| CN111025251B (en) * | 2019-11-22 | 2022-12-27 | 中国电子科技集团公司第二十研究所 | Multi-target composite detection method based on dynamic programming |
| CN110888126B (en) * | 2019-12-06 | 2023-01-17 | 西北工业大学 | Unmanned ship information perception system data comprehensive processing method based on multi-source sensor |
| CN112986947B (en) * | 2021-04-13 | 2021-07-23 | 南京雷电信息技术有限公司 | Machine learning-based trace point filtering processing method |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013167476A (en) * | 2012-02-14 | 2013-08-29 | Nec Corp | Target tracking processor and target tracking processing method |
| CN102608590B (en) * | 2012-03-21 | 2013-07-17 | 电子科技大学 | Coherent integration method based on dynamic programming and back image projection algorithm |
| CN102621543B (en) * | 2012-04-02 | 2015-02-25 | 中国人民解放军海军航空工程学院 | Dim target track-before-detect method based on particle filter algorithm and track management |
| CN103439697B (en) * | 2013-08-23 | 2015-05-27 | 西安电子科技大学 | Target detection method based on dynamic programming |
| CN103513244B (en) * | 2013-10-17 | 2016-03-30 | 电子科技大学 | A kind of multi-frame phase coherence accumulation target tracking-before-detecting method based on dynamic programming |
| CN104007422B (en) * | 2014-05-21 | 2016-06-15 | 西安电子科技大学 | Method is followed the tracks of based on before the multiple Likelihood ration test of dynamic programming |
| CN104035084B (en) * | 2014-06-24 | 2016-04-20 | 电子科技大学 | Tracking before a kind of dynamic programming for non-homogeneous clutter background detects |
| CN104391293A (en) * | 2014-11-07 | 2015-03-04 | 电子科技大学 | Pre-detection tracking method for space-time adaptive processing radar |
-
2015
- 2015-04-09 CN CN201510164452.8A patent/CN104793200B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN104793200A (en) | 2015-07-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104793200B (en) | Dynamic planning track-before-detect method based on iterative processing | |
| He et al. | Bounding box regression with uncertainty for accurate object detection | |
| CN108027972B (en) | System and method for object tracking | |
| He et al. | Infrared target tracking via weighted correlation filter | |
| CN106971401B (en) | Multi-target tracking device and method | |
| CN110766724A (en) | Target tracking network training and tracking method and device, electronic equipment and medium | |
| CN104865570B (en) | Tracking before a kind of quick Dynamic Programming detection | |
| US10592786B2 (en) | Generating labeled data for deep object tracking | |
| CN106934324A (en) | Based on the radar data correlating methods for simplifying many hypothesis algorithms | |
| KR101885839B1 (en) | System and Method for Key point Selecting for Object Tracking | |
| Zhang et al. | A GM-PHD algorithm for multiple target tracking based on false alarm detection with irregular window | |
| CN110376581B (en) | Explicit multi-target tracking method based on Gaussian mixture probability hypothesis density filter | |
| CN110187335A (en) | Tracking before being detected for the particle filter with discontinuous characteristic target | |
| CN111914878A (en) | Feature point tracking training and tracking method and device, electronic equipment and storage medium | |
| CN111027555B (en) | License plate recognition method and device and electronic equipment | |
| CN110766725B (en) | Template image updating method and device, target tracking method and device, electronic equipment and medium | |
| CN109919043B (en) | Pedestrian tracking method, device and equipment | |
| CN110276788B (en) | Method and apparatus for infrared imaging seeker target tracking | |
| Zhang et al. | Adaptive Gaussian mixture probability hypothesis density for tracking multiple targets | |
| CN108845299B (en) | Multi-sensor multi-frame joint detection algorithm based on posterior information fusion | |
| CN113192110A (en) | Multi-target tracking method, device, equipment and storage medium | |
| CN109284707B (en) | Moving target detection method and device | |
| Chen et al. | Fast compressive tracking combined with Kalman filter | |
| CN110031797B (en) | Detection and tracking method for passive sensing system on target with discontinuous characteristic | |
| CN104182990B (en) | A kind of Realtime sequence images motion target area acquisition methods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |