CN104766334A - Infrared weak and small target detection and tracking method and device - Google Patents
Infrared weak and small target detection and tracking method and device Download PDFInfo
- Publication number
- CN104766334A CN104766334A CN201510190521.2A CN201510190521A CN104766334A CN 104766334 A CN104766334 A CN 104766334A CN 201510190521 A CN201510190521 A CN 201510190521A CN 104766334 A CN104766334 A CN 104766334A
- Authority
- CN
- China
- Prior art keywords
- target
- infrared image
- information
- image
- cphd
- 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.)
- Granted
Links
Landscapes
- Closed-Circuit Television Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an infrared weak and small target detection and tracking method. According to an improved four-order partial differential equation method, an original infrared image is processed to obtain an infrared image with background suppression and target enhancement achieved, the position information and number information of candidate targets in the obtained infrared image are extracted according to a block adaptive threshold segmentation method, finally multi-target state and number estimation is carried out on the extracted position information and number information of the candidate targets according to Gaussian Mixture Cardinalized Probability Hypothesis Density (GM-CPHD) filter, and the states and number of multiple infrared weak and small targets are accurately and stably estimated through the GM-CPHD filter. The invention further discloses an infrared weak and small target detection and tracking device. Through the method and device, implementation is easy, the effect is obviously superior to that of a traditional background suppression method, the traditional multi-target tracking data association problem is avoided, and the multi-target states and number changing along with time can be estimated in time more stably.
Description
Technical field
The invention belongs to infrared image processing and target following technical field, be specifically related to a kind of small IR targets detection tracking and device thereof.
Background technology
Infrared imagery technique is by the infrared emanation work of receiving target, have completely passive, be easy to hidden, operating distance is far away, can the advantage such as work double tides, be widely used in multiple military affairs and the civil areas such as infrared precise guidance, early warning, video monitoring, search and tracking.The various application of infrared imagery technique be unable to do without the support of high performance infrared target detection and tracking technique, therefore the object detecting and tracking technology studying infrared imaging has great importance, and especially the detection and tracking of Infrared Target are the heat subjects of Recent study always.
When infrared eye operating distance is comparatively far away and working environment is complicated and changeable, the background of infrared image presents the space distribution of non-stationary, as the cloudy background risen and fallen, bright earth background and various clutters etc., residing for the target such as aircraft, guided missile wherein as plane only accounting for several pixel, and lack the information such as shape, texture, often be submerged in complicated background, and then present the low feature of contrast.And the background of infrared image sequence is time dependent, in actual combat, the difficulty of this type of target of detecting and tracking is very large.
Through finding the retrieval of existing technical literature and patent, at present for the method for pre-service many employings morphologic filtering of complex background infrared image, for No. 201210163140.1 patents, in the Image semantic classification step of this patent of invention, the method of morphologic filtering is adopted to obtain background image, obtain containing noisy background suppress image with original image subtracting background image, morphologic filtering is a kind of method of background forecast, it can realize Background suppression, but target can not be strengthened simultaneously, and in target tracking stage, the method adopted based on filtering and data correlation at present more, as data association algorithm comprises arest neighbors data correlation, probabilistic data association, JPDA, multiple hypotheis tracking, probability multiple hypotheis tracking etc., filtering algorithm comprises the Kalman filtering being applicable to linear Gauss model, and the EKF be applicable under non-linear Gauss conditions improved, Unscented kalman filtering and quadrature Kalman filter, be applicable to the particle filter etc. of nonlinear and non-Gaussian environment in addition.As No. 201210275678.1 patents, this patent of invention adopts the method for multiple hypotheis tracking data correlation to realize the Search/Track of infrared small object, data correlation is the core of traditional multi-object tracking method, it determines the performance of algorithm to a great extent, and when clutter data increase increase with target numbers time, the problem of shot array can be there is, under complex background, when the number of target changes in time, exist and detect uncertain problem, and then the uncertainty of data correlation can be caused.
Summary of the invention
In view of this, fundamental purpose of the present invention is to provide a kind of small IR targets detection tracking and device thereof.
For achieving the above object, technical scheme of the present invention is achieved in that
The embodiment of the present invention provides a kind of small IR targets detection tracking, the method is: carry out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved to original infrared image, positional information and the information of number of candidate target in the infrared image of described acquisition is extracted according to the method for block adaptive Threshold segmentation, gesture probability hypothesis density (GM-CPHD) wave filter finally realized according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
In such scheme, the described fourth order PDEs method according to improving carries out processing the infrared image after obtaining background suppress and targets improvement to original infrared image, be specially: establish k to represent the frame number of infrared image sequence, the size of infrared image is M × N, initial time k=1; First, read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
in, original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
In such scheme, the described method according to block adaptive Threshold segmentation extracts positional information and the information of number of candidate target in the infrared image of described acquisition, be specially: the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, the method according to mirror image is expanded original infrared image; For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
In such scheme, described gesture probability hypothesis density (GM-CPHD) wave filter finally realized according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter, be specially: the positional information of the candidate target of described extraction and information of number are carried out recurrence in the measurement feeding GM-CPHD wave filter of present frame.
In such scheme, the described positional information of the candidate target of described extraction and information of number feeding in GM-CPHD wave filter as the measurement of present frame carries out recurrence, realizes especially by following steps:
Step 501: the probability distribution of the state average of the target of previous frame survival and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Step 502: structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
Step 503: establish present frame to have Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
Step 504: the gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Step 505: extract multiobject state in the gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Step 506: judge whether present frame has newborn target, if had, then sends into the state of newborn target and number in the GM-CPHD prediction of next frame.
The embodiment of the present invention also provides a kind of small IR targets detection tracking means, and this device comprises: enhancement unit, extraction unit, filter element;
Described enhancement unit, for carrying out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved to original infrared image;
Described extraction unit, for extract described acquisition according to the method for block adaptive Threshold segmentation infrared image in the positional information of candidate target and information of number;
Described filter element, gesture probability hypothesis density (GM-CPHD) wave filter for realizing according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, is realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
In such scheme, described enhancement unit, specifically for the frame number establishing k to represent infrared image sequence, the size of infrared image is M × N, initial time k=1; Read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
in, original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
In such scheme, described extraction unit, specifically for the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, the method according to mirror image is expanded original infrared image; For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
In such scheme, described filter element, to send in GM-CPHD wave filter as the measurement of present frame specifically for the positional information of the candidate target using described extraction and information of number and carries out recurrence.
In such scheme, described filter element, the probability distribution specifically for the state average of the target of surviving to previous frame and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
If present frame has Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
The gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Extract multiobject state in gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Judge whether present frame has newborn target, if had, then the state of newborn target and number are sent in the GM-CPHD prediction of next frame.
Compared with prior art, beneficial effect of the present invention:
The present invention adopts the fourth order PDEs of improvement to realize suppression and the targets improvement of infrared image complex background, algorithm is simple, be easy to realize and successful is better than traditional background suppression method, GM-CPHD wave filter is utilized to carry out target following, avoid the data correlation problem of traditional multiple target tracking, and compared to GM-PHD wave filter, the time dependent multiple goal state of real-time estimation that can be more stable and number, can be widely used in the infrared small object tracker of complex background.
Accompanying drawing explanation
Fig. 1 is schematic flow sheet of the present invention;
Fig. 2 is the schematic flow sheet of GM-CPHD wave filter recurrence of the present invention.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.
The embodiment of the present invention provides a kind of small IR targets detection tracking, and as shown in Figure 1, the method is realized by following steps:
Step 101: original infrared image is carried out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved.
Concrete, if k represents the frame number of infrared image sequence, the size of infrared image is M × N, initial time k=1; First, read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
in, original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
(1) the fourth order PDEs principle, improved
Will detect infrared small object under complex background, first need Background suppression clutter, reduce false-alarm number, next needs to increase Weak target intensity, to improve contrast, and then improves verification and measurement ratio.Fourth order PDEs to be applied to small IR targets detection field, first need to reset coefficient of diffusion
value, wherein u represents the strength function of infrared image.If the coefficient of diffusion that behalf is new, then s arrange as shown in table 1.
Table 1, according to the feature of infrared image zones of different, adjusts the value of s (x)
According to the setting of new coefficient of diffusion, adopt the diffusivity expression that new, as (1) formula
If the infrared image after background suppress and targets improvement by
represent, then
can be expressed by (2) formula.
In formula
be the second order difference of image intensity value, (3) formula can be expressed as
In formula, i, j represent pixel present position in the picture.
(2), background suppress and targets improvement is realized according to the fourth order PDEs improved.
A. when certain area grayscale of image is uniformly distributed or the distribution of gray scale constant gradient, this region is likely background, at this moment according to (3) formula,
level off to 0, can be obtained by table one
and then by (2) Shi Ke get
namely background is suppressed or remove.
B. when certain zonule grey scale change in image is violent, probably there is interested target in this region, at this moment according to (3) formula, and can this region
and then by (2) Shi Ke get
therefore the gray-scale value in this region can be increased nearly 16N doubly, and namely target is enhanced.
C. in the edge contour region of image, also meet
therefore
and then by (2) Shi Ke get
therefore the gray scale of image border contour area can be increased nearly 4N doubly, and namely edge contour can by the enhancing of less degree.
Step 102: positional information and the information of number of extracting candidate target in the infrared image of described acquisition according to the method for block adaptive Threshold segmentation.
Concrete, the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, method according to mirror image is expanded original infrared image, namely M aliquant 50, N aliquant 50 time, if m1 is
integral part, n1 is
integral part, then expansion after image size be 50 × (m1+1) × (n1+1); For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
Step 103: gesture probability hypothesis density (GM-CPHD) wave filter finally realized according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, is realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
Concrete, the positional information of the candidate target of described extraction and information of number are carried out recurrence in the measurement feeding GM-CPHD wave filter of present frame.
The recurrence concrete steps of described GM-CPHD wave filter:
If the stochastic finite collection of any time k multiple goal state and measurement can be described by (5), (6) two formulas
Wherein, X in (5) formula
k-1represent the multiple goal state set in k-1 moment, S
k|k-1represent and to be survived target stochastic finite collection by the k-1 moment, Γ
krepresent the target stochastic finite collection of k moment new life, K in (6) formula
krepresent the clutter stochastic finite collection observed, Θ
kx () represents the measurement generated by the target in k moment.
If the upper limit of target numbers is Nmax in infrared image, and pruning threshold, merging threshold value and estimation threshold value are given.
The described positional information of the candidate target of described extraction and information of number feeding in GM-CPHD wave filter as the measurement of present frame carries out recurrence, as shown in Figure 2, realizes especially by following steps:
Step 201: the probability distribution of the state average of the target of previous frame survival and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Concrete, according to
And w
k|k-1=p
s× w
k-1to dbjective state average m in present frame infrared image
k|k-1, weight w
k|k-1and covariance P
k|k-1row prediction, wherein p
sfor the probability that the target of previous frame is still survived at present frame, predict according to the target numbers of GM-CPHD filtered target number Probability distribution prediction formula to present frame, obtain the number of targets probability distribution p predicted
k|k-1, then according to formula n
k|k-1=sum ([0:Nmax] × p
k|k-1) obtain the target numbers n of predicted current frame
k|k-1, regard each target predicted as a gauss component.
Step 202: structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
Concrete, construct gain K, the new breath covariance S required for GM-CPHD filtering renewal according to kalman filter method and upgrade covariance P
k|k.
Step 203: establish present frame to have Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
Concrete, using the candidate target number of adaptive threshold fuzziness link gained and state as the measured value of present frame, adopt the gauss component of the method for Kalman filtering to prediction to upgrade.If present frame has Z measuring value, then altogether obtain m after having upgraded
k|k-1+ Z × m
k|k-1individual gauss component.Each gauss component is containing the dbjective state average m upgraded
k|k, weight w
k|k, covariance P
k|k, according to the target numbers probability distribution p of GM-CPHD filtered target number more new formula and prediction
k|k-1calculate the target numbers probability distribution p upgraded
k|k.
Step 204: the gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Concrete, the gauss component that weights after GM-CPHD filtering renewal are less than pruning threshold is directly rejected, after beta pruning completes, select the gauss component of a maximum weight, travel through in remaining gauss component, if the mahalanobis distance between certain gauss component and gauss component of maximum weight is less than merging threshold value, then the gauss component of this gauss component and maximum weight is merged.One takes turns after merging completes, and repeats above-mentioned steps, until all gauss components are merged in remaining gauss component.
Step 205: extract multiobject state in the gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Concrete, weights are greater than and estimate that the gauss component of threshold value exports as dbjective state, and according to formula n
k|k=sum ([0:N_max] × p
k|k) calculate the target numbers of renewal.
Step 206: judge whether present frame has newborn target, if had, then sends into the state of newborn target and number in the GM-CPHD prediction of next frame.
The embodiment of the present invention also provides a kind of small IR targets detection tracking means, and this device comprises: enhancement unit, extraction unit, filter element;
Described enhancement unit, for carrying out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved to original infrared image;
Described extraction unit, for extract described acquisition according to the method for block adaptive Threshold segmentation infrared image in the positional information of candidate target and information of number;
Described filter element, gesture probability hypothesis density (GM-CPHD) wave filter for realizing according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, is realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
Described enhancement unit, specifically for the frame number establishing k to represent infrared image sequence, the size of infrared image is M × N, initial time k=1; Read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
Described extraction unit, specifically for the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, the method according to mirror image is expanded original infrared image; For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
Described filter element, to send in GM-CPHD wave filter as the measurement of present frame specifically for the positional information of the candidate target using described extraction and information of number and carries out recurrence.
Described filter element, the probability distribution specifically for the state average of the target of surviving to previous frame and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
If present frame has Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
The gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Extract multiobject state in gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Judge whether present frame has newborn target, if had, then the state of newborn target and number are sent in the GM-CPHD prediction of next frame.
The above, be only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.
Claims (10)
1. a small IR targets detection tracking, it is characterized in that, the method is: carry out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved to original infrared image, positional information and the information of number of candidate target in the infrared image of described acquisition is extracted according to the method for block adaptive Threshold segmentation, gesture probability hypothesis density (GM-CPHD) wave filter finally realized according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
2. small IR targets detection tracking according to claim 1, it is characterized in that, the described fourth order PDEs method according to improving carries out processing the infrared image after obtaining background suppress and targets improvement to original infrared image, be specially: establish k to represent the frame number of infrared image sequence, the size of infrared image is M × N, initial time k=1; First, read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
in, original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
3. small IR targets detection tracking according to claim 1, it is characterized in that, the described method according to block adaptive Threshold segmentation extracts positional information and the information of number of candidate target in the infrared image of described acquisition, be specially: the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, the method according to mirror image is expanded original infrared image; For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
4. small IR targets detection tracking according to claim 1, it is characterized in that, described gesture probability hypothesis density (GM-CPHD) wave filter finally realized according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter, be specially: the positional information of the candidate target of described extraction and information of number are carried out recurrence in the measurement feeding GM-CPHD wave filter of present frame.
5. small IR targets detection tracking according to claim 4, it is characterized in that, the described positional information of the candidate target of described extraction and information of number feeding in GM-CPHD wave filter as the measurement of present frame carries out recurrence, realizes especially by following steps:
Step 501: the probability distribution of the state average of the target of previous frame survival and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Step 502: structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
Step 503: establish present frame to have Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
Step 504: the gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Step 505: extract multiobject state in the gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Step 506: judge whether present frame has newborn target, if had, then sends into the state of newborn target and number in the GM-CPHD prediction of next frame.
6. a small IR targets detection tracking means, is characterized in that, this device comprises: enhancement unit, extraction unit, filter element;
Described enhancement unit, for carrying out processing the infrared image after obtaining background suppress and targets improvement according to the fourth order PDEs method improved to original infrared image;
Described extraction unit, for extract described acquisition according to the method for block adaptive Threshold segmentation infrared image in the positional information of candidate target and information of number;
Described filter element, gesture probability hypothesis density (GM-CPHD) wave filter for realizing according to Gaussian Mixture carries out multiple goal state and number estimation to the positional information of the candidate target extracted and information of number, is realized the state of multiple infrared small object and the precise and stable estimation of number by GM-CPHD wave filter.
7. small IR targets detection tracking means according to claim 6, is characterized in that: described enhancement unit, and specifically for the frame number establishing k to represent infrared image sequence, the size of infrared image is M × N, initial time k=1; Read the kth frame in infrared image, adopt the fourth order PDEs improved to process kth two field picture, by formula (1):
coefficient of diffusion s (x) defined brings formula (2) into:
original infrared image is processed, obtains the infrared image after background suppress and targets improvement.
8. small IR targets detection tracking means according to claim 6, it is characterized in that: described extraction unit, specifically for the infrared image after described acquisition background suppress and targets improvement is divided into 50 fritters, when the height of infrared image and width do not meet the multiple of 50, the method according to mirror image is expanded original infrared image; For each fritter after division, according to formula (4): point block threshold value=(block image average+12 × (block image variance)) × (adaptive thresholding value coefficient), determine segmentation threshold, pixel assignment gray scale being greater than threshold value is 255, the pixel assignment being less than threshold value is 0, after adaptive threshold fuzziness completes, it is capable that 1 of intercepting expanded images arrives M, 1 to N arranges as image after Threshold segmentation, according to positional information and the information of number of image determination candidate target after segmentation.
9. small IR targets detection tracking means according to claim 6, it is characterized in that: described filter element, to send in GM-CPHD wave filter as the measurement of present frame specifically for the positional information of the candidate target using described extraction and information of number and carry out recurrence.
10. small IR targets detection tracking means according to claim 9, it is characterized in that, described filter element, probability distribution specifically for the state average of the target of surviving to previous frame and the newborn target of present frame, weights, covariance and number is predicted, obtains the target mean m predicted
k|k-1, weight w
k|k-1, covariance P
k|k-1with the target numbers n of prediction
k|k-1and the probability distribution p of target numbers
k|k-1;
Structure target information upgrades required composition, namely constructs gain K, new breath covariance S and upgrades covariance P
k|k;
If present frame has Z measuring value, then altogether obtain m
k|k-1+ Z × m
k|k-1individual gauss component, adopts Kalman filtering to upgrade each gauss component, obtains the state average m upgraded
k|k, weight w
k|kwith covariance P
k|k, and the probability distribution p to target of prediction number
k|k-1upgrade, obtain the target numbers probability distribution p upgraded
k|k;
The gauss component according to the method for beta pruning weights being less than threshold value is rejected, and in the gauss component after beta pruning, state difference is less than the gauss component merging threshold value and merges;
Extract multiobject state in gauss component after beta pruning merges and calculate the target numbers upgraded, exporting as last filtering;
Judge whether present frame has newborn target, if had, then the state of newborn target and number are sent in the GM-CPHD prediction of next frame.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510190521.2A CN104766334B (en) | 2015-04-21 | 2015-04-21 | Small IR targets detection tracking and its device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510190521.2A CN104766334B (en) | 2015-04-21 | 2015-04-21 | Small IR targets detection tracking and its device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104766334A true CN104766334A (en) | 2015-07-08 |
CN104766334B CN104766334B (en) | 2017-12-29 |
Family
ID=53648139
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510190521.2A Active CN104766334B (en) | 2015-04-21 | 2015-04-21 | Small IR targets detection tracking and its device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104766334B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105469090A (en) * | 2015-11-19 | 2016-04-06 | 南京航空航天大学 | Frequency-domain-residual-error-based small target detection method and apparatus in infrared image |
CN106096604A (en) * | 2016-06-02 | 2016-11-09 | 西安电子科技大学昆山创新研究院 | Multi-spectrum fusion detection method based on unmanned platform |
CN106709497A (en) * | 2016-11-16 | 2017-05-24 | 北京理工大学 | PCNN-based infrared motion weak target detection method |
CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
CN107369164A (en) * | 2017-06-20 | 2017-11-21 | 成都中昊英孚科技有限公司 | A kind of tracking of infrared small object |
CN108665435A (en) * | 2018-01-08 | 2018-10-16 | 西安电子科技大学 | The multispectral section of infrared image background suppressing method based on topology-Tu Qie fusion optimizations |
CN109493365A (en) * | 2018-10-11 | 2019-03-19 | 中国科学院上海技术物理研究所 | A kind of tracking of Weak target |
CN109902578A (en) * | 2019-01-25 | 2019-06-18 | 南京理工大学 | A kind of infrared target detection and tracking |
CN110298865A (en) * | 2019-05-22 | 2019-10-01 | 西华大学 | The space-based Celestial Background small point target tracking of cluster device is separated based on threshold value |
CN110490904A (en) * | 2019-08-12 | 2019-11-22 | 中国科学院光电技术研究所 | A kind of Dim targets detection and tracking |
CN111428573A (en) * | 2020-03-02 | 2020-07-17 | 南京莱斯电子设备有限公司 | Infrared weak and small target detection false alarm suppression method under complex background |
CN111598922A (en) * | 2020-04-13 | 2020-08-28 | 华中光电技术研究所(中国船舶重工集团公司第七一七研究所) | Infrared weak and small target self-adaptive tracking algorithm switching method and system under sea-air background |
CN115311470A (en) * | 2022-09-28 | 2022-11-08 | 北京万龙精益科技有限公司 | Infrared small target real-time detection and tracking method of adaptive block matching filtering |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110040525A1 (en) * | 2009-08-17 | 2011-02-17 | Lockheed Martin Corporation | Method and system for calculating elementary symmetric functions of subsets of a set |
CN102542540A (en) * | 2011-12-31 | 2012-07-04 | 华中科技大学 | Method for inhibiting infrared image background based on PDE (Partial Differential Equation) |
CN104408482A (en) * | 2014-12-08 | 2015-03-11 | 电子科技大学 | Detecting method for high-resolution SAR (Synthetic Aperture Radar) image object |
CN104616299A (en) * | 2015-01-30 | 2015-05-13 | 南京邮电大学 | Method for detecting weak and small target based on space-time partial differential equation |
-
2015
- 2015-04-21 CN CN201510190521.2A patent/CN104766334B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110040525A1 (en) * | 2009-08-17 | 2011-02-17 | Lockheed Martin Corporation | Method and system for calculating elementary symmetric functions of subsets of a set |
CN102542540A (en) * | 2011-12-31 | 2012-07-04 | 华中科技大学 | Method for inhibiting infrared image background based on PDE (Partial Differential Equation) |
CN104408482A (en) * | 2014-12-08 | 2015-03-11 | 电子科技大学 | Detecting method for high-resolution SAR (Synthetic Aperture Radar) image object |
CN104616299A (en) * | 2015-01-30 | 2015-05-13 | 南京邮电大学 | Method for detecting weak and small target based on space-time partial differential equation |
Non-Patent Citations (4)
Title |
---|
李严 等: "基于四阶偏微分方程的图像去噪方法实现", 《现代电子技术》 * |
欧阳成 等: "一种改进的CPHD多目标跟踪算法", 《电子与信息学报》 * |
章涛 等: "自适应门限GM_CPHD多目标跟踪算法", 《数据采集与处理》 * |
龚俊亮: "红外弱小多目标实时检测跟踪技术研究", 《中国博士学位论文全文库 信息科技辑》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105469090B (en) * | 2015-11-19 | 2018-12-18 | 南京航空航天大学 | Small target detecting method and device in infrared image based on frequency-domain residual |
CN105469090A (en) * | 2015-11-19 | 2016-04-06 | 南京航空航天大学 | Frequency-domain-residual-error-based small target detection method and apparatus in infrared image |
CN106096604A (en) * | 2016-06-02 | 2016-11-09 | 西安电子科技大学昆山创新研究院 | Multi-spectrum fusion detection method based on unmanned platform |
CN106709497A (en) * | 2016-11-16 | 2017-05-24 | 北京理工大学 | PCNN-based infrared motion weak target detection method |
CN107256560A (en) * | 2017-05-16 | 2017-10-17 | 北京环境特性研究所 | A kind of method for detecting infrared puniness target and its system |
CN107256560B (en) * | 2017-05-16 | 2020-02-14 | 北京环境特性研究所 | Infrared weak and small target detection method and system thereof |
CN107369164A (en) * | 2017-06-20 | 2017-11-21 | 成都中昊英孚科技有限公司 | A kind of tracking of infrared small object |
CN107369164B (en) * | 2017-06-20 | 2020-05-22 | 成都中昊英孚科技有限公司 | Infrared weak and small target tracking method |
CN108665435A (en) * | 2018-01-08 | 2018-10-16 | 西安电子科技大学 | The multispectral section of infrared image background suppressing method based on topology-Tu Qie fusion optimizations |
CN108665435B (en) * | 2018-01-08 | 2021-11-02 | 西安电子科技大学 | Multi-spectral-band infrared image background suppression method based on topology-graph cut fusion optimization |
CN109493365A (en) * | 2018-10-11 | 2019-03-19 | 中国科学院上海技术物理研究所 | A kind of tracking of Weak target |
CN109902578A (en) * | 2019-01-25 | 2019-06-18 | 南京理工大学 | A kind of infrared target detection and tracking |
CN110298865A (en) * | 2019-05-22 | 2019-10-01 | 西华大学 | The space-based Celestial Background small point target tracking of cluster device is separated based on threshold value |
CN110490904A (en) * | 2019-08-12 | 2019-11-22 | 中国科学院光电技术研究所 | A kind of Dim targets detection and tracking |
CN110490904B (en) * | 2019-08-12 | 2022-11-11 | 中国科学院光电技术研究所 | Weak and small target detection and tracking method |
CN111428573B (en) * | 2020-03-02 | 2021-06-25 | 南京莱斯电子设备有限公司 | Infrared weak and small target detection false alarm suppression method under complex background |
CN111428573A (en) * | 2020-03-02 | 2020-07-17 | 南京莱斯电子设备有限公司 | Infrared weak and small target detection false alarm suppression method under complex background |
CN111598922A (en) * | 2020-04-13 | 2020-08-28 | 华中光电技术研究所(中国船舶重工集团公司第七一七研究所) | Infrared weak and small target self-adaptive tracking algorithm switching method and system under sea-air background |
CN115311470A (en) * | 2022-09-28 | 2022-11-08 | 北京万龙精益科技有限公司 | Infrared small target real-time detection and tracking method of adaptive block matching filtering |
CN115311470B (en) * | 2022-09-28 | 2023-01-24 | 北京万龙精益科技有限公司 | Infrared small target real-time detection and tracking method of adaptive block matching filtering, system and device thereof and computer readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN104766334B (en) | 2017-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104766334A (en) | Infrared weak and small target detection and tracking method and device | |
He et al. | Infrared target tracking via weighted correlation filter | |
CN101930072B (en) | Multi-feature fusion based infrared small dim moving target track starting method | |
CN104766320B (en) | Many Bernoulli Jacob under thresholding is measured filter Faint target detection and tracking | |
CN104851097B (en) | The multichannel SAR GMTI methods aided in based on target shape and shade | |
CN102722706B (en) | Particle filter-based infrared small dim target detecting and tracking method and device | |
CN101482969B (en) | SAR image speckle filtering method based on identical particle computation | |
CN101964060B (en) | SAR variant target identification method based on local textural feature | |
CN101867699A (en) | Real-time tracking method of nonspecific target based on partitioning | |
CN110555870B (en) | DCF tracking confidence evaluation and classifier updating method based on neural network | |
CN104021559A (en) | Image registration method based on mutual information and Harris corner point detection | |
CN107742113B (en) | One kind being based on the posterior SAR image complex target detection method of destination number | |
CN106023134A (en) | Automatic grain boundary extraction method for steel grain | |
CN104899892A (en) | Method for quickly extracting star points from star images | |
CN104156929A (en) | Infrared weak and small target background inhibiting method and device on basis of global filtering | |
Mitchell et al. | A semi-automatic approach for estimating near surface internal layers from snow radar imagery | |
Der et al. | Probe-based automatic target recognition in infrared imagery | |
Reichman et al. | Improvements to the Histogram of Oriented Gradient (HOG) prescreener for buried threat detection in ground penetrating radar data | |
CN105354824A (en) | Region extraction-based two-parameter constant false alarm detection method | |
CN104680536A (en) | Method for detecting SAR image change by utilizing improved non-local average algorithm | |
CN109461127B (en) | SAR image sparse regularization feature enhancement method with interpretation as purpose | |
CN108986139B (en) | Feature integration method with significance map for target tracking | |
CN104537384A (en) | SAR (synthetic aperture radar) target identification method combined with likelihood ratio decision | |
CN106056115A (en) | Infrared small target detection method under non-uniform background | |
Kapp et al. | Spatial verification of high-resolution ensemble precipitation forecasts using local wavelet spectra |
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 |