CN101339657B - TBD target discrimination method based on direction Histogram statistics - Google Patents
TBD target discrimination method based on direction Histogram statistics Download PDFInfo
- Publication number
- CN101339657B CN101339657B CN2008101177130A CN200810117713A CN101339657B CN 101339657 B CN101339657 B CN 101339657B CN 2008101177130 A CN2008101177130 A CN 2008101177130A CN 200810117713 A CN200810117713 A CN 200810117713A CN 101339657 B CN101339657 B CN 101339657B
- Authority
- CN
- China
- Prior art keywords
- target
- histogram statistics
- tbd
- angle
- statistics
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to a goal identification method of TBD-DP algorithm. The steps of the invention are that the target trajectory in the TBD-DP test results is vectorized; vector coordinates are corresponding to the position of image frames in targets; (2) based on the general direction which is decided by the target vectorization position in a start frame and a termination frame, the direction deviation angles of the targets between different the image frames are calculated; (3) the direction deviation angles carry out histogram statistics; (4) real targets or false alarms to be distinguished are determined according to the results of the histogram statistics. The invention uses the difference that the noise gives rise to the false alarms, which has strong randomicity on trajectory, while the real target trajectory approximates to a straight in the shortest time; based on the multi-level direction histogram statistics, the targets are distinguished, thus achieving the purpose of eliminating the false alarms caused by the noise in the test results. The simulation results also show that the histogram statistics method has better effect than a local maximum method used currently in the elimination of the false alarms caused by the noise.
Description
Technical field
The invention belongs to the preceding target discrimination method of following the tracks of (TBD, Track Before Detecting) algorithm of a kind of detection, be applied to little target detection and tracing process, belong to the aftertreatment field of little target detection and tracking.
Background technology
Weak target is the lower target of signal to noise ratio (S/N ratio).Along with low detectable technology rapid development of modern times and widespread use, making becomes one of important research content in radar and the optical imagery field to the detection of weak target and the research of method for early warning.In recent years, a lot of methods is being proposed aspect the detection of weak target, wherein aspect the Weak target of detection and tracking low signal-to-noise ratio be a kind of very effective method based on track algorithm before the detection of dynamic programming (TBD-DP, Track Before Detecting Based on Dynamic Programming).The point that this method comprises the hypothesis path between frame and frame does not almost have the relevant treatment of information loss, and through the accumulation of number frame, after the track of target was estimated, the flight path of testing result and target was announced simultaneously.But, can there be a lot of false-alarms among the result that TBD-DP detects owing to very noisy and target diffusion influence in the dynamic programming process.Be to eliminate false-alarm, method commonly used at present has based on the removing method of the track of overlapping with based on the removing method of local extremum (LEV, local extreme value).The false-alarm that these two kinds of methods cause for target diffusion is effectively, but the identification result of the false-alarm that causes for very noisy is unsatisfactory.
Summary of the invention
The objective of the invention is provides a kind of TBD-DP target discrimination method that can effectively differentiate the false target that noise causes at based on the false-alarm that is caused by noise in the track algorithm before the detection of dynamic programming, to improve the performance of TBD-DP algorithm.
The present invention has followed following technical scheme: a kind of TBD-DP algorithm target discrimination method based on direction Histogram statistics may further comprise the steps:
(1) for carrying out vectorized process based on the target trajectory in the track algorithm TBD-DP testing result before the detection of dynamic programming, the phasor coordinate correspondence the position of target in two field picture;
(2) general direction with target vector determining positions in start frame and the abort frame is a benchmark, calculates target at the deviation in driction angle of different images interframe, and computing formula is dir
d=dir
I, j-dir
t, dir
I, j=angle (v
i-v
j), v
iAnd v
jBe respectively the position of same target in two two field pictures and be respectively, angle () is for getting angle, and the scope of value is-π～π that i-j＞0 is the direction exponent number;
(3) the direction fleet angle is carried out Histogram statistics;
(4) result according to Histogram statistics judges be real goal or false-alarm to be identified.
Described step (2) is calculated in the target direction fleet angle process, has used multistage deviation in driction angle, and for avoiding the influence of discretize, the selection of exponent number need limit, usually greater than 2 rank.
Described step (3) is divided the statistics interval according to total fleet angle number when carrying out Histogram statistics, each interval on average has 2-5 fleet angle approximately.
Behind the travel direction Histogram statistics, the deviation in driction angle that falls into 0 district is added up, definite symmetrical region [a, a] that adopted in 0 district, the span of a can be limited to 1 times interval the arriving between π/4 of statistics.
Described step (4) need be determined decision threshold when judging real goal or false-alarm, thresholding is according to the certain proportion setting of total fleet angle number, when scale-up factor can be set to that fleet angle accumulative total surpasses thresholding in 0.5～0.75,0 district, target to be identified is a real goal, otherwise is false-alarm.
The present invention's advantage compared with prior art is:
(1) the present invention utilizes noise to cause that false-alarm has very strong randomness on track, and the real goal track is approximately this difference of straight line in the short period, carries out target based on multistage direction Histogram statistics and differentiates.And then reaching the purpose of eliminating the false-alarm that causes by noise in the testing result, simulation result shows that also the direction Histogram statistics method has better effect than the local maximum method of using at present aspect the false-alarm that noise causes eliminating.
(2) discrimination method of the present invention not only can effectively be removed the false-alarm that noise causes in the TBD-DP algorithm, and does not relate to complex calculations in the implementation procedure, and is simple.
Description of drawings
Fig. 1 is the process flow diagram that the present invention is based on the TBD target discrimination method of direction Histogram statistics;
Fig. 2 is the fleet angle histogram of exemplary trajectory of the present invention.
Wherein: the 101:TBD-DP method is carried out the result of target detection
102: the target location vector quantization
103: calculate the target trajectory fleet angle
104: the fleet angle Histogram statistics
105: judge target genuine-fake
201: the track fleet angle histogram of the false target that noise causes
202: real goal track fleet angle histogram.
Embodiment
It is the subsequent treatment of target detection process that target is differentiated, its objective is the false target of eliminating in the testing result.For the TBD-DP algorithm, owing to be characterized in that after the track of target was estimated, the flight path of testing result and target announced that simultaneously therefore, the discrimination process of testing result can make full use of the trace information of target through the accumulation of number frame.Adopt the TBD-DP method to carry out target detection, and adopt discrimination method of the present invention to carry out flow process that target differentiates as shown in Figure 1.Target discrimination process has wherein comprised target location vector quantization 102, calculating target trajectory fleet angle 103, fleet angle Histogram statistics 104 and has judged 105 4 steps of target genuine-fake, TBD testing process 101 itself does not belong to the target discrimination process, but differentiates testing result and the target trajectory information of providing for target.Discrimination method among the present invention is to carry out on the basis that the TBD-DP algorithm has provided testing result and target trajectory, has comprised following 4 steps.
1, target location vector quantization
Behind TBD-DP, the track correspondence of target the position in every two field picture, and the position of hypothetical target in a certain two field picture is that r is capable, c row, and then the target location can be expressed as vector form and is
v＝c+jr(1)
2, target trajectory divergence angle calculations
The position of same target in two two field pictures is respectively v
iAnd v
j, then objective definition is at v
iAnd v
jBetween corresponding deflection be
dir
i，j＝angle(v
i-v
j)(2)
In the formula, angle () is for getting angle, and the scope of value is-π～π that i-j=1 for the direction exponent number, is worked as in i-j＞0, then is two adjacent two field pictures.
When the i value is a last frame, j=1 obtains dir
I, jBe total direction of motion, be designated as dir
t, the target trajectory that so just can obtain corresponding i-j rank departs from dir
dAngle
dir
d＝dir
i，j-dir
i(3)
In the direction calculating of different rank, because the discretize of target state, the situation that moving target corresponding same position and dbjective state sudden change in multiple image can occur, therefore when the travel direction Histogram statistics, can limit the direction exponent number, as greater than 2 rank, in application, can limit the direction exponent number and be 3 rank, 4 rank even high-order more.
3, fleet angle Histogram statistics
Since the frame number difference of image accumulation in the TBD process, the fleet angle number difference that when carrying out divergence angle calculations, obtains, but satisfy following relation:
N
k＝F-k (4)
Be k fleet angle exponent number in the formula, F is accumulation frame number, N
kBe the fleet angle number that calculates.With the accumulation of 15 frames is example, and 1 rank fleet angle number is that 14,2 rank fleet angle numbers are 13,3 rank fleet angle numbers are 12,, consider the influence of discretize, generally get greater than 2 rank, when getting 2,3,4,5,6 rank, total fleet angle number is 55, and total fleet angle number is very limited, therefore in Histogram statistics, statistics is interval too much unsuitable, can divide according to the amount of each interval average about 2-5 fleet angle.
4, target genuine-fake is judged
The false target that causes for noise, track has randomness, the track deviation angle shows even distribution, and target trajectory has definite direction of motion, in the short period of time, track is approximately straight line, and the track deviation angle concentrates near 0, and typical noise causes that the distribution of track fleet angle of false-alarm and real goal is respectively shown among Fig. 2 201 and 202.By the trajectory direction fleet angle is carried out Histogram statistics, and the fleet angle that falls into 0 near zone (can be called for short 0 zone) being counted, thought real goal when count value surpasses when setting thresholding, is false-alarm less than what set thresholding.
Carrying out target genuine-fake judges as follows:
Here relate to determining of two parameters, i.e. 0 zone and judgement threshold, 0 zone is meant a near symmetrical region 0, can be expressed as [a, a], the span of a can be limited to 1 times interval π/4 of arriving of statistics; Judgement threshold is set to the certain proportion of total fleet angle usually, as 0.5～0.75 times fleet angle number.
For verifying the validity of this discrimination method, carried out emulation experiment, noise is caused the identification result of false-alarm for checking, target is not set in emulation, 15 two field pictures that only comprise noise are carried out TBD-DP to be detected, adopt direction Histogram statistics method and local maximum method to differentiate respectively to testing result, the target numbers (false-alarm) before and after differentiating is as shown in table 1.Wherein carried out 3 rank and 4 rank directional statistics in the direction Histogram statistics, 0 regional extent is set to
, the counting threshold setting for always count 65%.The Monte Carlo experiment number of times is 200 times.
Number of targets before and after table 1 is differentiated
| Before the discriminating | The Histogram statistics method | The local maximum method | |
| The false-alarm number | 1474 | 247(16.8％) | 637(43.2％) |
False-alarm targets number before and after differentiating from table 1 as can be seen, the residue false-alarm targets only is 16.8% after the employing direction Histogram statistics method, and adopt the local maximum method to handle back residue false-alarm ratio is 43.2%, therefore, for the false-alarm targets that is caused by noise fully, the direction Histogram statistics method is more effective than local maximum method.
In a word, the present invention utilizes noise to cause that false-alarm has very strong randomness on track, and the real goal track is approximately this difference of straight line in the short period, carrying out target based on multistage direction Histogram statistics differentiates, and then reaching the purpose of eliminating the false-alarm that causes by noise in the testing result, simulation result shows that also the direction Histogram statistics method has better effect than the local maximum method of using at present aspect the false-alarm that noise causes eliminating.
The content that is not described in detail in the instructions of the present invention belongs to this area professional and technical personnel's known prior art.
Although disclose most preferred embodiment of the present invention and accompanying drawing for the purpose of illustration, it will be appreciated by those skilled in the art that: without departing from the spirit and scope of the invention and the appended claims, various replacements, variation and modification all are possible.Therefore, the present invention should not be limited to most preferred embodiment and the disclosed content of accompanying drawing.
Claims (6)
1. TBD-DP algorithm target discrimination method based on direction Histogram statistics is characterized in that step is as follows:
(1) for carrying out vectorized process based on the target trajectory in the track algorithm TBD-DP testing result before the detection of dynamic programming, the phasor coordinate correspondence the position of target in two field picture;
(2) general direction with target vector determining positions in start frame and the abort frame is a benchmark, calculates target at the deviation in driction angle of different images interframe, and computing formula is dir
d=dir
I, j-dir
t, dir
I, j=angle (v
i-v
j), v
iAnd v
jBe respectively the position of same target in two two field pictures; Angle () is for getting angle, and the scope of value is-π～π, i-j＞0, and i-j is the direction exponent number; Dir
tBe total direction of motion, dir
dBe the deviation in driction angle; Dir
I, jBe the deflection between same target i two field picture position and the j two field picture position;
(3) the direction fleet angle is carried out Histogram statistics;
(4) result according to Histogram statistics judges be real goal or false-alarm to be identified.
2. according to the TBD-DP algorithm target discrimination method based on direction Histogram statistics of claim 1, it is characterized in that: described step (2) is calculated in the target direction fleet angle process, used multistage deviation in driction angle, for avoiding the influence of discretize, the selection of direction exponent number need limit, greater than 2 rank.
3. according to the TBD-DP algorithm target discrimination method based on direction Histogram statistics of claim 1, it is characterized in that: described step (3) is when carrying out Histogram statistics, divide the statistics interval according to total fleet angle number, on average there is 2-5 fleet angle in each statistics interval.
4. according to the TBD-DP algorithm target discrimination method based on direction Histogram statistics of claim 1 or 3, it is characterized in that: behind the travel direction Histogram statistics, the deviation in driction angle that falls into 0 district is added up, definite symmetrical region [a that adopted in 0 district, a], the span of a is limited to 1 times interval the arriving between π/4 of statistics.
5. according to the TBD-DP algorithm target discrimination method of claim 1 based on direction Histogram statistics, it is characterized in that: described step (4) need be determined decision threshold when judging real goal or false-alarm, thresholding is according to the certain proportion setting of total fleet angle number, scale-up factor is set to 0.5～0.75, when fleet angle accumulative total surpasses thresholding in 0 district, target to be identified is a real goal, otherwise is false-alarm.
6. according to the TBD-DP algorithm target discrimination method of claim 1 based on direction Histogram statistics, it is characterized in that: described step (1) vectorized process is: the position of target in a certain two field picture is that r is capable, the c row, and then the vector form of target location is:
v＝c+jr。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101177130A CN101339657B (en) | 2008-08-04 | 2008-08-04 | TBD target discrimination method based on direction Histogram statistics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101177130A CN101339657B (en) | 2008-08-04 | 2008-08-04 | TBD target discrimination method based on direction Histogram statistics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101339657A CN101339657A (en) | 2009-01-07 |
| CN101339657B true CN101339657B (en) | 2011-04-20 |
Family
ID=40213717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101177130A Expired - Fee Related CN101339657B (en) | 2008-08-04 | 2008-08-04 | TBD target discrimination method based on direction Histogram statistics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101339657B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102313884B (en) * | 2010-06-29 | 2013-02-13 | 电子科技大学 | Target track-before-detect (TBD) method based on multi-frame coherent integration |
| 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 |
| CN104881561B (en) * | 2014-08-22 | 2017-09-29 | 中国科学院沈阳自动化研究所 | Tracking before a kind of detection of the multi-Dimensional parameters based on Hough transform |
| CN104282023A (en) * | 2014-09-27 | 2015-01-14 | 郑敏 | Target trajectory estimation method based on preprocessing |
-
2008
- 2008-08-04 CN CN2008101177130A patent/CN101339657B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN101339657A (en) | 2009-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103530893B (en) | Based on the foreground detection method of background subtraction and movable information under camera shake scene | |
| CN109859239B (en) | A kind of method and apparatus of target tracking | |
| CN109636829A (en) | A kind of multi-object tracking method based on semantic information and scene information | |
| CN105405151A (en) | Anti-occlusion target tracking method based on particle filtering and weighting Surf | |
| CN103143158B (en) | Method for improving snooker hit rate | |
| CN108765470A (en) | One kind being directed to the improved KCF track algorithms of target occlusion | |
| CN101339657B (en) | TBD target discrimination method based on direction Histogram statistics | |
| CN103455820A (en) | Method and system for detecting and tracking vehicle based on machine vision technology | |
| CN102789568A (en) | Gesture identification method based on depth information | |
| CN110208789B (en) | Multi-frame combined target track identification method | |
| CN102810250A (en) | Video based multi-vehicle traffic information detection method | |
| CN102722702B (en) | Multiple feature fusion based particle filter video object tracking method | |
| CN109993769A (en) | A kind of multiple-target system of deep learning SSD algorithm combination KCF algorithm | |
| CN103996051A (en) | Method for automatically detecting abnormal behaviors of video moving object based on change of movement features | |
| CN108596157B (en) | Crowd disturbance scene detection method and system based on motion detection | |
| CN104239843B (en) | Positioning method and device for face feature points | |
| CN106408563B (en) | A kind of snow noise detection method based on the coefficient of variation | |
| CN103617633B (en) | Cascading identification method for spatial infrared target | |
| Li et al. | Multiple weld seam laser vision recognition method based on the IPCE algorithm | |
| CN108830204A (en) | The method for detecting abnormality in the monitor video of target | |
| CN106291498A (en) | A kind of detecting and tracking combined optimization method based on particle filter | |
| Almajai et al. | Anomaly detection and knowledge transfer in automatic sports video annotation | |
| CN103996207A (en) | Object tracking method | |
| Huang et al. | An algorithm of the target detection and tracking of the video | |
| CN111027558A (en) | Method for recognizing type and license plate of violation vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110420 Termination date: 20110804 |