CN101567087A - Method for detecting and tracking small and weak target of infrared sequence image under complex sky background - Google Patents
Method for detecting and tracking small and weak target of infrared sequence image under complex sky background Download PDFInfo
- Publication number
- CN101567087A CN101567087A CNA2009100854699A CN200910085469A CN101567087A CN 101567087 A CN101567087 A CN 101567087A CN A2009100854699 A CNA2009100854699 A CN A2009100854699A CN 200910085469 A CN200910085469 A CN 200910085469A CN 101567087 A CN101567087 A CN 101567087A
- Authority
- CN
- China
- Prior art keywords
- target
- image
- track
- mrow
- background
- 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
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 42
- 238000001514 detection method Methods 0.000 claims abstract description 32
- 230000008030 elimination Effects 0.000 claims abstract description 14
- 238000003379 elimination reaction Methods 0.000 claims abstract description 14
- 238000009825 accumulation Methods 0.000 claims abstract description 13
- 230000002401 inhibitory effect Effects 0.000 claims abstract description 5
- 238000006073 displacement reaction Methods 0.000 claims description 15
- 230000001629 suppression Effects 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 230000002829 reductive effect Effects 0.000 claims description 4
- 230000036961 partial effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 abstract 1
- 230000010339 dilation Effects 0.000 abstract 1
- 230000007774 longterm Effects 0.000 description 9
- 230000003287 optical effect Effects 0.000 description 9
- 230000003044 adaptive effect Effects 0.000 description 8
- 238000012804 iterative process Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 230000000877 morphologic effect Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 230000005484 gravity Effects 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 229910003460 diamond Inorganic materials 0.000 description 2
- 239000010432 diamond Substances 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 240000005373 Panax quinquefolius Species 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000005206 flow analysis Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a method for detecting and tracking a small and weak target of an infrared sequence image under complex sky background, which comprises the following steps of: (1) adopting a self-adaptive background eliminating algorithm which combines with the small and weak target characteristic of an image and Top-hat conversion for conducting self-adaptive background elimination and inhibiting the influence of complex background and part of noise on the small and weak target detection; (2) using an iterative threshold algorithm which is suitable for a dim-background image for dividing the possible target area in the dim-background image after background elimination; (3) constructing a dilation accumulation algorithm according to the motion law of the target in a plurality of continuous frames, and distinguishing possible false-alarm target areas; (4) using CB algorithm and logic and operation based on contour structure elements for inhibiting the false-alarm targets; (5) taking the areas in which the false-alarm targets are inhibited as possible target area which is then compared with the track prediction result in the step (7) so as to realize the updating of track; (6) conducting track judgment; and (7) conducting track prediction.
Description
Technical Field
The invention relates to a method for detecting and tracking small and weak targets of an infrared sequence image under a complex sky background, belongs to the field of digital image processing, and mainly relates to mathematical morphology and infrared image and target detection and tracking technologies. The method is widely applied to various target recognition or navigation systems based on infrared images.
Background
The research of the weak and small target detection and tracking technology under the complex sky background has important application in civil, aerospace and military (see the literature: Zhang-Tian sequence, etc. motion weak point target detection based on three-dimensional width search direction filtering. mode identification promulgation, 2007, 28 (2): 246-253.(Tianxu Zhang, Meng Li, Zhengrong Zuo, Weidong Yang, Xiechang Sun, moving point target detection with same-dimensional with-to-exact search directional filters, 2007, 28 (2): 246-253.)) is an important research field of the key technology such as image processing, target identification, navigation and guidance. Due to the fact that the imaging distance is long and the influence of a complex sky background is caused, the signal to noise ratio of a small target in an infrared image is low, the small target is located in the complex background, the motion speed is unknown, and exact shape information does not exist, so that the difficulty of detecting and tracking the infrared weak and small target is greatly increased. At this time, in order to effectively improve the effectiveness of detecting and tracking the infrared weak and small target under the complex background, it is necessary to effectively utilize the characteristics of the target to enhance the detectability of the target, simplify the difficulty of confirming the target track during target tracking, and improve the effectiveness of target detection and tracking.
In order to effectively improve the effectiveness of detecting and tracking weak and small targets under a complex infrared background, various methods for effectively utilizing the characteristics of the targets are proposed. These methods include:
(1) methods based on filtering (see [1] Dasipander et al. [1] Desmopander S.D., RondaM.H.Chan Er, V., Ph... Max-Mean and Max-media filters for detection of Small-target. SPIE, Signal and Data Processing of Small Targets 1999, 74-83.) [2] Seff et al.. Optimization of Point target tracking filters [ American society of Electrical and electronics Engineers ] Aero-electronics, 2000, 36 (1): 15-25.([2] Cafer C.E., Silverman J., Mooney J.M. Optimization of target tracking, and [2] electronics J.E., Silverman J., Objective J.M. Optimization of target tracking, 25 ] detection of Small target Adaptation of Electronic Systems [1] Across.3, 25 ] Observation of Electronic Systems [ 25 ] Objections, Objective enhancement of Electronic Systems [ 25 ] Objections, Objective Objections, and Data Processing of Small target Adaptation of Objections [ 25 ] Objections [3, Objections ] Objections [ 25 ] Objections ] Adaptation, Max-media engineering, Max Institute of engineers image processing journal, 1997, 6 (3): 383-
The basic idea of the method is to construct a corresponding filter according to the target characteristics, and filter the image through the filter so as to filter out a complex background, thereby achieving the purpose of improving the signal-to-noise ratio of the image and simplifying the difficulty of target detection. The method has poor effect when the image signal-to-noise ratio is low, or related parameters are extremely difficult to select, or the algorithm is extremely complex and difficult to apply in real time.
(2) Wavelet-based methods (see [1] Chengxingclock, Sunworth. energy feature-based image target detection. Infrared and laser engineering, 2001, 30 (1): 30-32.[2] Marlet. Multi-resolution Signal decomposition Theory: wavelet representation. American society of Electrical and electronics Engineers model analysis and machine Intelligent Collection, 1989, 11 (7): 674 693.([2] Mallat, S.G.. Theory for multiplexing decision: the detection of weak targets in the multi-scale transform-based front view infrared image. International society of optical engineering, pami, 1989, 11 (7): 674 and 693.) [3] Grace et al.. 1994, 2262: 62-71.([3] Gregs, D.J., Deutput of IR target detection: 62: 71.) (SPP. simulation, Sp. S.D.J.). J.: 2262: 62-71)
The method mainly utilizes a multi-scale analysis method of wavelets to distinguish texture backgrounds of images and interested target areas. The image information generated by the wavelet on different scales provides characteristic information which can distinguish the target from the complex background, namely, when the target image is analyzed by the wavelet, the characteristics of the target and the background on different scales of the wavelet are obviously different. Accordingly, various wavelet-based methods are better applied to the detection and tracking of small targets with complex backgrounds. Wavelet based detection algorithms cannot be applied well to situations where the image signal-to-noise ratio is low.
(3) Morphology-based methods (see [1] treatment in board, sheng, Zhou dawn, etc.; application of wavelet transformation and mathematical morphology in infrared image detection. laser and infrared, 2003, 33 (2): 94-97 [2] white phase, etc.; small target enhancement in infrared based on novel top-hat transformation, journal of electronic Imaging, 2008, 17 (3): 030501.([2] Bai Xiangzhi, Zhou Fugen, Xieyongchun.New class of top-hat transformation to electron Imaging, 2008, 17 (3): 030501.)
The main operation involved in this type of method is morphological Top-hat transformation. The main idea of using Top-hat transformation to detect small targets is to estimate the background by Top-hat transformation and subtract the background from the original image to obtain the target. However, this method is poor in fluctuating background and low in signal-to-noise ratio.
(4) Optical flow-based methods (see [1] Bibo et al. calculation of Optical flow based on extended constraints. American institute of Electrical and electronics Engineers image processing journal, 1996, 5 (5): 720- > 739.([1] Bimbo A.D., Nesi P., join L.C.S.. Optical flow calculation extended constraints. IEEE trans.IP., 1996, 5 (5): 720- > 739.) [2] Lianchun, Sun-Co 25035, Pan nationwide, et al. A Rapid Algorithm for division of eastern targets in complex scenarios. the university of Qinghua university newspaper (Nature edition), 1997, 37 (8): 59-64.)
The method can determine the number of moving objects, the moving speed, the object distance and the surface structure of the objects through optical flow analysis. However, due to the aperture problem, the complexity and the calculation amount of the optical flow algorithm, when the optical flow is applied to real-time detection and tracking of an actual target, the algorithm efficiency, the algorithm evaluation, the optical flow robust estimation and the optical flow fast estimation still need to be studied intensively.
(5) Methods based on the Multistage Hypothesis (MHT) (see, Rede et al, a Multi-target tracking Algorithm, American institute of Electrical and electronics Engineers, avionics systems, J.1979, 24: 843 & 854.(Reid D.B.. An Algorithm for tracking multiple targets. IEEE Trans AES, 1979, 24: 843 & 854.))
The method organizes a plurality of possible target tracks in a tree structure, and performs hypothesis test on the trees reaching each frame of image in the sequence at the same time to remove the trees which do not pass the test at any time, thereby reducing the operation amount and the storage amount. However, in the complex background target detection and tracking, the amount of memory and calculation required for implementing the tracking algorithm based on the multi-stage assumption is too large and the situation description (including the number of targets and the motion state) at a certain moment is determined, so that the application of the tracking method based on the multi-stage assumption is limited.
(6) A method based on data interconnection (see document [1] Basalon et al. [1] Classification-aided multiframe data interconnection Tracking. [1] Bar-Shalom Y.,. Kirubaajan T.,. Gokberk C.. Tracking with classification-aided multiframe data association. IEEE Transactions on Aerosa and Electronic System, 2005, 41 (3): 868 877.) [2] spring, Yexining, Zhang Honghuai. generalized probability data association algorithm, 2005, 33 (3): 467 472 [3] Basalon. multiobjective multi-sensor Tracking principle and method, Marksai, Yak et al. [3] Shaak-1990 ] Barsalon et 1990. Barsalon. multistep et 1990. Australi, Australian and Electronic systems, 2005, 41 (3): 868 877.) 2] spring, Yexining, Massachusen
The method is better applied to multi-target tracking. However, when the target number and the effective measurement number are increased, the number of the feasible joint events is increased exponentially, so that the calculated amount of the data interconnection algorithm is increased exponentially, even a combined explosion phenomenon occurs, and meanwhile, the complexity of constructing the feasible joint events is increased, so that the method can obtain a better effect by corresponding design according to a specific application object.
(7) Motion correspondence-based methods (see [1] D. et al.. Motion correspondence establishment based on extended transient range. Artificial intelligence, 2003, 145: 227-
The method directly detects and tracks the targets according to the actual motion rules of various targets. However, under the condition of more false alarm targets, the method has longer operation time and is difficult to meet the application of a real-time system, so that the method needs further research in the practical application.
In addition, other methods are proposed, such as a method based on curve evolution, a method based on a neural network, a method based on a random field, and the like, but these methods are specific algorithms for specific targets, and have poor universality.
Disclosure of Invention
The infrared weak and small target detection and tracking under the complex sky background must be combined with the target detection technology of single-frame images and the target tracking technology of multi-frame images, and the real target under the complex background is detected and tracked through the gray distribution characteristics of the target in the single-frame images and the motion characteristics of the target in the multi-frame images. Meanwhile, the situation that the target is blocked may exist in the moving process of the target, and the target loss situation caused by blocking can be compensated through prediction under the simple blocking situation. However, in the case of long-term occlusion, prediction of an object and re-detection after occlusion become difficult problems.
In order to solve the problems and make up for the defects of the traditional method, the invention provides a method for detecting and tracking the small and weak targets of the infrared sequence image under the complex sky background, the effectiveness of target detection and tracking under the complex background is improved by effectively utilizing the characteristics of the targets in the single-frame image and the motion characteristics of the targets in the multi-frame image, the problems of target prediction and re-detection after occlusion under the long-term occlusion condition are solved, and an effective solution is provided for the practical application problem in the practical system.
To achieve this, the technical solution of the present invention is to detect possible targets in an image by using adaptive background rejection and iterative threshold segmentation. The self-adaptive background elimination can effectively detect the target under the complex background, and can also detect the target which is shielded for a long time again. And then, inhibiting a false alarm target generated by a complex background and noise by utilizing the expansion accumulation and the false alarm target, and reducing the difficulty of follow-up target tracking. And finally, confirming the final real target by utilizing a target tracking technology. In target tracking, the invention constructs a new multi-frame prediction algorithm for solving target prediction under the condition of long-term occlusion and provides a reference for target tracking when a target is occluded for a long time.
The invention is characterized by comprising the following specific steps:
(1) self-adaptive background elimination: and self-adaptive background elimination is carried out by adopting a self-adaptive background elimination algorithm combining the image small and weak target characteristics and Top-hat transformation, so that the influence of a complex background and partial noise on the detection of the small and weak target is inhibited. The target re-detection after long-term occlusion is the re-detection of the target, and can be realized by using a self-adaptive background elimination algorithm.
(2) Iterative threshold segmentation: giving an initial value, dividing the image into a target type and a background type, and performing iteration by using the initial value and two types of gray values in the image to make the initial value and the two types of gray values in the image converge to a final threshold value;
(3) expansion accumulation: and constructing an expansion accumulation algorithm according to the motion rule of the target in continuous multiframes, and performing morphological expansion and accumulation on certain continuous images. Due to the continuity of the real infrared target motion, the real target area in the image after expansion accumulation is necessarily a larger area, and other false alarm target areas are smaller, so that possible false alarm target areas can be distinguished;
(4) false alarm target suppression: and filtering out a possible false alarm target area generated by expansion accumulation by using CB morphological operation and logic and operation based on the contour structural elements, and inhibiting the false alarm target.
(5) Track updating: taking the area subjected to false alarm target suppression as a possible target area, performing distance comparison with the result of the track prediction in the step (7), and replacing the predicted target area position with the possible target area position to realize track updating;
(6) and (3) track judgment: and judging the track of the real target by using the continuous motion information of the target. If a certain track is not updated in a certain frame image, the track loses the target once in the frame. Counting the number of times that the target in a certain track is lost, if the number of times that the target is lost is more than the preset number of times, the track is formed by the false alarm target, and the track is removed; otherwise, the track is reserved, and the target corresponding to the track is a real target;
(7) and (3) track prediction: in order to predict the track and solve the problem of predicting the target under the condition of long-term occlusion, the position of the target of the next frame is calculated by using the average value of the motion displacement of the previous frames, the possible position of each track in the image of the next frame is predicted by effectively using the motion information of the previous frames, and the influence of the prediction deviation caused by the fluctuation of the displacement of a few frames on the prediction result is reduced.
Wherein, the trajectory prediction in step (7) is to calculate the position of the next frame target by using the mean value of the motion displacements of the previous frames, so as to reduce the influence of the prediction deviation caused by the fluctuation of the displacements of a few frames on the prediction result, and the specific method is as follows:
the number of image frames used for prediction is set as nP, wherein the displacement of a target corresponding to a certain track of the ith frame (i is more than or equal to 1 and less than or equal to nP-1) is as follows:
mdi＝||Pi-Pi1||，
then the possible positions of the target corresponding to the track in the nP-th frame are:
where round is rounding and rounding, and nP is typically odd.
The invention has the advantages and effects that: the self-adaptive background elimination is utilized to suppress the influence of complex background and noise on the detection of the weak and small targets, so that the detection of the targets and the re-detection of the targets after long-term shielding are realized; the expansion accumulation and the false alarm target inhibition which effectively utilize the motion characteristics of the target inhibit the generation of partial false alarm targets and reduce the difficulty of tracking and confirming the target by the subsequent target; effective track prediction, updating and judgment algorithms are constructed by utilizing the continuous motion rule of the real target, so that the effective detection of the target and the prediction of the target position under the long-term shielding condition are realized. Meanwhile, the track prediction uses the mean value of the motion displacement of the previous frames to calculate the position of the target of the next frame, so that the influence of prediction deviation caused by the fluctuation of a few frames of displacement on a prediction result is reduced, the target prediction under the long-term shielding condition is effectively solved, and the target tracking is simplified. The method can effectively inhibit the influence of complex sky background and noise on the detection and tracking of the infrared dim target, can be widely applied to various target identification and navigation systems based on infrared images, and has wide market prospect and application value.
Drawings
Fig. 1 is a schematic block diagram of a method for detecting and tracking a small and weak target of an infrared sequence image under a complex sky background according to the present invention.
Fig. 2 shows the experimental results of the present invention on a small target image sequence of an airplane comprising 174 frames of images, and the small rectangular boxes in the figure mark the results tracked by the detection of the present invention. Graphs (a), (b), (c), (d), (e) and (f) are experimental results for frames 13, 21, 25, 38, 128 and 138 in the sequence, respectively.
Detailed Description
For better understanding of the technical solutions of the present invention, the following further describes embodiments of the present invention with reference to the accompanying drawings. The flow of the invention is shown in fig. 1, and the specific implementation details of each part are as follows:
1. adaptive background rejection
The adaptive morphological background rejection (AMCE) algorithm flow is as follows.
Let the length and width of the image f be Wide and Height respectively:
the first step is as follows: calculating R ═ CBCB(f)，
And
respectively carrying out morphological expansion and corrosion operation;
the second step is that: setting i to be 0;
the third step: calculating jv (i) ═ maxwi-minwiWherein maxwiAnd minwiIs of size Lw×LhAnd a rectangular window w centered on pixel iiMaximum and minimum values of the middle pixel gray;
the fourth step: the image background corresponding to pixel i is estimated as follows:
the fifth step: if i is less than Wide × Height, i is i +1, and the third step is carried out;
otherwise, turning to the sixth step;
and a sixth step: rejecting the estimated image background using:
I＝f-BG。
the AMCE algorithm flow described above removes the estimated image background BG from f,the target is retained while the image background is suppressed, thereby achieving enhancement of the target. W in AMCEiShould be smaller than the size of the possible target area and can be obtained by a priori knowledge or experiment of the target.
2. Iterative threshold segmentation
Under the ideal condition, after the estimated image background is subtracted from the original image for adaptive background elimination, a target image only containing a target point and image noise is left, but in an actual condition, due to inaccuracy of a background estimation algorithm on background estimation, the target image also contains residual image background. The remaining image background is a large number of low gray value pixels, making the target image a dark background image. The histogram of this type of image has no distinct peaks, but the number of low pixel value pixels is the largest.
Aiming at the characteristics of the target image with low gray value, the image can be divided into two types of target and background, an initial value is given, and iteration is carried out by utilizing the initial value and the two types of gray values in the image so as to lead the initial value and the two types of gray values in the image to be converged to a final threshold value. The process is adaptive to the dark background image in various situations.
The adaptive iteration threshold steps are as follows:
(1) taking the average value of the whole image after the self-adaptive background elimination as the initial value of Threshold
Wherein N is the number of pixels, ri＝(xi，yi) Is the pixel coordinate of the image, I (r)i) Is riAnd the pixel value of the corresponding pixel point.
(2) Dividing the image into two categories by Threshold, taking the category with the pixel value larger than Threshold as the target, and marking the mean value as meanupThe class with pixel values less than Threshold is the background, and the mean value is denoted meandownThe new Threshold is calculated as follows:
Threshold＝(1-1/f(x))×meandown+(1/f(x))×meanup，
wherein,
x＝meanup-meandown，
(3) If the value of Threshold no longer changes, the iteration is aborted, otherwise (2) is returned.
In the above iterative process, in the calculation formula of the Threshold, f (x) is selected as follows:
the value range of f (x) should be (1, ∞)
This allows the value of Threshold not to exceed the gray scale range.
(x) is a non-decreasing function of x
This will allow f (x) to have an auto-regulating effect on Threshold for the following reasons:
when Threshold is in the high gray value region, x is meanup-meandownLarger, when (1-1/f (x)) is larger and (1/f (x)) is smaller, mean is increased in the calculation of Threshold value in step (2)downThe specific gravity of (a) reduces Threshold; in contrast, when Threshold is in a low gray value region, x ═ meanup-meandownSmaller, with (1-1/f (x)) smaller and (1/f (x)) larger, the calculation of Threshold in step (2) will reduce meandownThe specific gravity of (A) increases Threshold; and mean when Threshold reaches a reasonable ThresholdupAnd meandownThere will be no change in the value of (a), (b), (c), and (d), and Threshold will settle at a reasonable Threshold. Thus, the algorithm automatically adjusts the Threshold value to the final Threshold value as the iterative process progresses.
It is obvious that Then the algorithm is droppedBecomes a commonly used iterative threshold algorithm.
From the above analysis, it can be seen that the values of f (x) are different in different iterative processes, and thus mean is adjusted in the processupAnd meandownThe adjustment coefficients of (a) are automatically adjusted according to Threshold, and the adjustment coefficients are different corresponding to different thresholds, which can be regarded as multi-scale adjustment, and this is fundamentally different from the conventional iterative Threshold algorithm, and the multi-scale adjustment process will make the thresholds converge on more reasonable thresholds.
(x) the curvature of f should be small
This allows the adjustment of f (x) not to be excessive to reduce the number of iterations.
Convergence of iterative procedure
Assuming Threshold as a reasonable Threshold for the image to be segmented, the gray level mean values of the corresponding target and background are meanupAnd meandownTheir difference is x ═ meanup-meandownMean since Threshold is now a reasonable Thresholdup、meandownAnd x will not change.
If the initial Threshold corresponding to a certain step in the iterative process is Threshold1 and Threshold1 < Threshold, the mean value of the image gray levels of the corresponding target and the background and the difference value thereof are meanup1、meandown1 and x1, then, knowing from the characteristics of the dark background image that x1 is less than x, i.e. Δ x1 is x-x1 > 0, step (2) will increase mean after one iterationupThe specific gravity of 1 increases Threshold1 to Threshold2, and the mean value of the image gray levels of the corresponding target and the background and the difference value thereof are meanup2、meandown2 and x2, mean from the characteristics of the dark background imageup2＞meanup1、meandown2＞meandown1, and meanup2-meanup1＞meandown2-meandown1, i.e. x2 > x1, so Δ x2 ═ x-x2 < Δ x1, which means mean as the iterative process progressesup-meandownThe difference Δ x between x corresponding to the reasonable threshold will gradually decrease, and when Δ x is zero, the corresponding threshold is the reasonable threshold, at which point the iterative process is terminated.
Similarly, if the initial Threshold corresponding to a step in the iterative process is Threshold1 and Threshold1 > Thershold, as the iterative process progresses, Threshold1 will also converge to a reasonable Threshold.
Experiments have shown that the iteration process can be terminated at a reasonable threshold value, typically performed less than 10 times.
From the above conditions, f (x) can be constructed as follows:
f(x)＝log10(10+αx)，
the alpha value is a constant and is different for different images. The smaller the mean value of the image, the smaller the value of α, generally between (0, 10). And obtaining a possible target image IPtar by using a result image after Threshold binarization self-adaptive background elimination.
3. Expansion summation
The motion rule of the real infrared target in the continuous multiframes is continuous, and the position of the false alarm target in the continuous multiframes is random. Thus, the inflation accumulation algorithm can be used to mark possible false alarm targets. The detailed steps of the algorithm are as follows:
setting the current frame as the ith frame binary target image IPtariThe number of frames of images to be accumulated is nS.
The first step is as follows: taking the j-th i-nS frame binary image, and initializing all pixels in the accumulation result image AddImap to be 0;
the second step is that: using the structure element B for the j frame image1Is subjected to swelling, B1The shape of (B) is usually circular1The size of (A) is usually 1-3;
the third step: adding the result image of the second step with AddImap directly;
the fourth step: if j is less than i, j is i +1, the second step is carried out,
otherwise, the algorithm terminates;
due to the continuity of the real infrared target motion, the real target area in the image after expansion accumulation is necessarily a larger area, and other false alarm target areas are smaller. Thus, this attribute may be used to suppress the false alarm target.
4. False alarm target suppression
From the results of the expansion summation, the size of the real target area is larger than that of the other areas. Therefore, the region with smaller size can be removed by using morphological operation, and the current frame image can be reconstructed by using logical AND operation, so that the false alarm target in the current frame image can be well suppressed. The detailed flow of the algorithm is as follows:
the first step is as follows: with CBPerforming operation processing AddImap, and removing an area with a smaller size; cB(f)＝min{f，CBCB(f)}；
The second step is that: IPtar (linear time warping) of original binary image of current frameiAnd CBAnd performing logic and operation on the calculated AddIimage, and reconstructing a current frame binary target image.
In the first step of the algorithm, C is adoptedBThe operation is mainly due to CBThe operation has the characteristic of protecting image details, so CBThe operation does not damage the real target area while removing the false alarm target. CBThe size of the structural elements used in the operation should be smaller than the real target area, and the specific size can be obtained by the prior knowledge. Since the false alarm target has no fixed shape, the shape of the structural element may be in the form of a conventional diamond or rectangle. The present invention suggests using a diamond with a radius of 3.
Due to the suppression of the false alarm target in AddImap, the AND operation in the second step of the algorithm can reconstruct the real target area in the current frame and remove part of the false alarm target, so as to achieve the purpose of suppressing the false alarm target. The resulting image after this step is denoted as RT.
5. Track update
Due to the suppression and expansion accumulation of the adaptive background elimination on the complex background of the image and the suppression of the false alarm target by the false alarm target suppression, the number of the false alarm targets in the result image RT is greatly reduced, but a small number of the false alarm targets are still reserved. At this time, a target tracking technique capable of effectively utilizing target motion information is used to confirm a real target. Here, the purpose of the trajectory update is to compare the position of a possible target in the current frame of the existing prediction with the detected position of a possible target in the current frame, so as to update the target position of each trajectory in the current frame.
The detailed updating steps of a certain Trace when there are NR regions in the current frame RT are as follows:
the first step is as follows: taking the predicted position P of the current frame in TraceCurLet i equal to 0, i equal to or more than 0 and equal to or less than NR-1;
the second step is that: get the ith region P in RTi；
The third step: calculating PiAnd PCuA distance d betweeni；
The fourth step: if i is not more than NR-1, i is i +1, then the second step is carried out;
otherwise, turning to the fifth step;
the fifth step: find all diMinimum value of dj；
And a sixth step: if d isjIf not more than mindT, use PiUpdating P in TraceCu；
Otherwise, the Trace is not updated;
the mindT in the algorithm flow described above can generally be obtained from a priori knowledge. The algorithm ensures that the track is continuously updated according to the target detection condition of the current frame, and ensures that the information of continuous motion of the target can be effectively utilized.
6. Trajectory decision
The purpose of the track judgment is to judge whether a certain track is a motion track of a real target by utilizing the continuous motion information of the target so as to achieve the purpose of identifying the target. Since the previous steps effectively suppress false alarm targets, the use of simple motion information can effectively decide real target trajectories.
If the Trace is not updated in a certain frame of image, the Trace is indicated to lose the target once in the frame. If the times of the Trace losing the target exceed a preset threshold value nTL, the Trace is not the track of the real target; otherwise, Trace is the trajectory of the real target. nTL is a predetermined threshold, the more severe the image noise, the larger nTL.
The decision method is based on the following basic facts: if the target is a real target, due to the continuity of the target motion, the track corresponding to the target is necessarily continuous and the situation that the target is lost in more frames does not exist; and due to the randomness of the false alarm target in continuous multi-frame images, the corresponding track of the false alarm target is lost in more frames.
Meanwhile, due to effective suppression of the false alarm target, a few false alarm targets with long tracks can be formed. Therefore, the track of the real target can be found after the track judgment. The position of the point of the current frame corresponding to the track is the position of the real target in the current frame, so that the aim of identifying the real target is fulfilled.
7. Trajectory prediction
One of the objects of the present invention is to solve the multi-frame occlusion problem. When the target is blocked by the object, the target cannot be detected, and the position of the target must be predicted according to the motion rule of the target in order to still identify the possible position of the target. Meanwhile, even if the target is not shielded, the prediction of the position of the next frame of the target is beneficial to providing a reference for the update of the target position of the current frame in the track in the target update.
Linear prediction can be used for predicting the target track, but simple linear prediction is only a basic assumption that the target motion rule in a short term can be approximated to a straight line. Therefore, in case of long-term occlusion of the target, simple linear prediction is less suitable for trajectory prediction. The reasons for this result are: simple linear prediction usually predicts the motion of the next frame with the motion of the first few frames (typically 2-3 frames). While the motion of the first few frames may fluctuate significantly, the bias produced in the case of predicting fewer frames is also small. But such fluctuations may introduce large errors in the case of predicting multiple frames. Therefore, it is necessary to effectively use the motion information of the previous multiframe to predict the result of the future multiframe. Therefore, the invention provides a prediction algorithm based on multi-frame target displacement.
The number of image frames used for prediction is set as nP, wherein the displacement of a target corresponding to a certain track of the ith frame (i is more than or equal to 1 and less than or equal to nP-1) is as follows:
mdi＝||Pi-Pi-1||，
then the possible positions of the target corresponding to the track in the nP-th frame are:
where round is rounding and rounding, and nP is typically odd.
As can be seen from the above formula, the prediction algorithm uses the average value of the motion displacement of the previous frames to calculate the position of the target of the next frame, and reduces the influence of the prediction deviation caused by the fluctuation of the displacement of a few frames on the prediction result.
After the trajectory prediction, the result is sent to the trajectory update for updating the trajectory of the real target.
In order to show the effect of the invention, the effectiveness of the invention is verified by adopting the airplane small target image sequence containing 174 frames of images. The image background in the image sequence is complex, the image noise is serious, and the target is shielded by the building background for a long time. The experimental results are shown in fig. 2, which shows the images at frames 13, 21, 25, 38, 128 and 138 of the image sequence. From the results, it can be seen that, although the image background is complex, the adaptive background estimation better suppresses the influence of the complex background on the effect of the present invention. Meanwhile, a large number of false alarm points may be generated by serious noise of the image, and the false alarm points are also inhibited, so that the method can obtain better effect. More importantly, the target in the image sequence is occluded for a long time, but the method still better predicts the possible position of the target and embodies the effectiveness of the invention.
Claims (2)
1. A method for detecting and tracking a small target of an infrared sequence image under a complex sky background is characterized by comprising the following steps: the method comprises the following specific steps:
(1) adopting a self-adaptive background elimination algorithm combining image small and weak target characteristics and Top-hat transformation to carry out self-adaptive background elimination, and inhibiting the influence of a complex background and partial noise on small and weak target detection;
(2) segmenting possible target areas in the dark background image after background removal by using an iterative threshold algorithm suitable for the dark background image;
(3) constructing an expansion accumulation algorithm according to the motion rule of the target in continuous multiframes, and distinguishing possible false alarm target areas;
(4) with C based on outline-structuring elementsBOperation and logical and operation suppress false alarm targets;
(5) comparing the area subjected to false alarm target suppression with the result of the track prediction in the step (7) as a possible target area to realize track updating;
(6) carrying out track judgment; if a certain track is not updated in a certain frame of image, the track loses the target once in the frame; counting the number of times of target loss in a certain track, if the number of times of target loss is more than the preset number of times, the track is formed by a false alarm target, and the track is removed; otherwise, the track is reserved, and the target corresponding to the track is a real target;
(7) predicting a track; the method effectively utilizes the motion information of a plurality of frames at the previous stage, and adopts a prediction algorithm based on the displacement of a plurality of frames of targets to predict the possible positions of targets in all tracks in the next frame of image.
2. The trajectory prediction in step (7) according to claim 1, characterized by: the method is characterized in that the average value of the motion displacement of a plurality of previous frames is used for calculating the position of a next frame target, so that the influence of prediction deviation caused by fluctuation of displacement of a few frames on a prediction result is reduced, and the specific method is as follows:
the number of image frames used for prediction is set as nP, wherein the displacement of a target corresponding to a certain track of the ith frame (i is more than or equal to 1 and less than or equal to nP-1) is as follows:
mdi＝||Pi-Pi-1||，
then the possible positions of the target corresponding to the track in the nP-th frame are:
where round is rounding and rounding, and nP is typically odd.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100854699A CN101567087B (en) | 2009-05-25 | 2009-05-25 | Method for detecting and tracking small and weak target of infrared sequence image under complex sky background |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100854699A CN101567087B (en) | 2009-05-25 | 2009-05-25 | Method for detecting and tracking small and weak target of infrared sequence image under complex sky background |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101567087A true CN101567087A (en) | 2009-10-28 |
| CN101567087B CN101567087B (en) | 2012-05-23 |
Family
ID=41283232
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009100854699A Expired - Fee Related CN101567087B (en) | 2009-05-25 | 2009-05-25 | Method for detecting and tracking small and weak target of infrared sequence image under complex sky background |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101567087B (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102005054A (en) * | 2010-11-24 | 2011-04-06 | 中国电子科技集团公司第二十八研究所 | Real-time infrared image target tracking method |
| CN102096925A (en) * | 2010-11-26 | 2011-06-15 | 中国科学院上海技术物理研究所 | Real-time closed loop predictive tracking method of maneuvering target |
| CN102184534A (en) * | 2011-05-25 | 2011-09-14 | 北京航空航天大学 | Method for image fusion by using multi-scale top-hat selective transform |
| CN103310419A (en) * | 2013-06-18 | 2013-09-18 | 深圳大学 | Small target image processing device |
| CN103335636A (en) * | 2013-05-31 | 2013-10-02 | 南京理工大学 | Detection method of small targets on ground |
| CN103617633A (en) * | 2013-11-26 | 2014-03-05 | 北京航空航天大学 | Cascading identification method for spatial infrared target |
| CN103679176A (en) * | 2013-12-30 | 2014-03-26 | 北京航空航天大学 | Linear feature extraction method utilizing multi-scale multi-constructing-element top-hat transformation |
| CN104103080A (en) * | 2014-07-02 | 2014-10-15 | 华中科技大学 | Method of small dim target detection under complicated background |
| CN104766100A (en) * | 2014-10-22 | 2015-07-08 | 中国人民解放军电子工程学院 | Infrared small target image background predicting method and device based on machine learning |
| CN105046713A (en) * | 2015-08-12 | 2015-11-11 | 北京航空航天大学 | Morphology-based robot star point segmentation method and FPGA realization method |
| CN105139432A (en) * | 2015-08-05 | 2015-12-09 | 西安电子科技大学 | Gaussian model based infrared small weak target image simulation method |
| CN105654516A (en) * | 2016-02-18 | 2016-06-08 | 西北工业大学 | Method for detecting small moving object on ground on basis of satellite image with target significance |
| CN106327522A (en) * | 2016-08-25 | 2017-01-11 | 上海航天控制技术研究所 | Infrared small target detection method based on multi-direction morphological filtering complex cloud background |
| CN107564018A (en) * | 2017-08-30 | 2018-01-09 | 北京航空航天大学 | It is a kind of to utilize the method for improving iterative algorithm extraction target image |
| CN107976685A (en) * | 2017-11-20 | 2018-05-01 | 北京航空航天大学 | A kind of infrared sensor indoor human body Target Tracking System based on Internet of Things |
| CN108805878A (en) * | 2018-05-22 | 2018-11-13 | 深圳腾视科技有限公司 | A kind of foreground object detection solution based on computer vision |
| CN109712169A (en) * | 2018-11-15 | 2019-05-03 | 上海卫星工程研究所 | Moving-target motion profile prediction technique and method for autonomous tracking based on EO-1 hyperion |
| CN109741396A (en) * | 2018-12-12 | 2019-05-10 | 天津津航技术物理研究所 | A kind of extremely small and weak infrared target detection method |
| CN109781142A (en) * | 2019-01-14 | 2019-05-21 | 上海卫星工程研究所 | Angle measurement is tested the speed Integrated Navigation Semi-physical Simulation verifying system and its test method |
| CN109859250A (en) * | 2018-11-20 | 2019-06-07 | 北京悦图遥感科技发展有限公司 | A kind of outer video multi-target detection of aviation red and tracking and device |
| CN109872345A (en) * | 2019-02-27 | 2019-06-11 | 中国科学院光电技术研究所 | A kind of monotrack method under dark background |
| CN109903305A (en) * | 2019-01-24 | 2019-06-18 | 天津国为信息技术有限公司 | Line style target impact point positioning method based on aerial three-dimensional localization |
| CN109978851A (en) * | 2019-03-22 | 2019-07-05 | 北京航空航天大学 | A kind of aerial weak moving target detection tracking of infrared video |
| CN110502968A (en) * | 2019-07-01 | 2019-11-26 | 西安理工大学 | The detection method of infrared small dim moving target based on tracing point space-time consistency |
| CN110728635A (en) * | 2019-09-10 | 2020-01-24 | 中国科学院上海技术物理研究所 | Contrast enhancement method for dim and weak target |
| CN111583283A (en) * | 2020-05-20 | 2020-08-25 | 北京字节跳动网络技术有限公司 | Image segmentation method, image segmentation device, electronic device, and medium |
| CN112183221A (en) * | 2020-09-04 | 2021-01-05 | 北京科技大学 | Semantic-based dynamic object self-adaptive trajectory prediction method |
| CN117095029A (en) * | 2023-08-22 | 2023-11-21 | 中国科学院空天信息创新研究院 | Method and device for detecting small target in air flight |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1274146C (en) * | 2002-10-10 | 2006-09-06 | 北京中星微电子有限公司 | Sports image detecting method |
| JP3970877B2 (en) * | 2004-12-02 | 2007-09-05 | 独立行政法人産業技術総合研究所 | Tracking device and tracking method |
| CN100565244C (en) * | 2007-10-31 | 2009-12-02 | 北京航空航天大学 | Multimode multi-target accurate tracking apparatus and method |
| CN101382997B (en) * | 2008-06-13 | 2011-06-22 | 青岛海信电子产业控股股份有限公司 | Vehicle detecting and tracking method and device at night |
| CN101354359B (en) * | 2008-09-04 | 2010-11-10 | 湖南大学 | Method for detecting, tracking and recognizing movement visible exogenous impurity in medicine liquid |
-
2009
- 2009-05-25 CN CN2009100854699A patent/CN101567087B/en not_active Expired - Fee Related
Cited By (42)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102005054A (en) * | 2010-11-24 | 2011-04-06 | 中国电子科技集团公司第二十八研究所 | Real-time infrared image target tracking method |
| CN102096925A (en) * | 2010-11-26 | 2011-06-15 | 中国科学院上海技术物理研究所 | Real-time closed loop predictive tracking method of maneuvering target |
| CN102184534A (en) * | 2011-05-25 | 2011-09-14 | 北京航空航天大学 | Method for image fusion by using multi-scale top-hat selective transform |
| CN103335636A (en) * | 2013-05-31 | 2013-10-02 | 南京理工大学 | Detection method of small targets on ground |
| CN103335636B (en) * | 2013-05-31 | 2015-07-22 | 南京理工大学 | Detection method of small targets on ground |
| CN103310419A (en) * | 2013-06-18 | 2013-09-18 | 深圳大学 | Small target image processing device |
| CN103310419B (en) * | 2013-06-18 | 2015-10-21 | 深圳大学 | A kind of Small object image processing apparatus |
| CN103617633A (en) * | 2013-11-26 | 2014-03-05 | 北京航空航天大学 | Cascading identification method for spatial infrared target |
| CN103617633B (en) * | 2013-11-26 | 2017-01-25 | 北京航空航天大学 | Cascading identification method for spatial infrared target |
| CN103679176A (en) * | 2013-12-30 | 2014-03-26 | 北京航空航天大学 | Linear feature extraction method utilizing multi-scale multi-constructing-element top-hat transformation |
| CN104103080B (en) * | 2014-07-02 | 2017-01-11 | 华中科技大学 | Method of small dim target detection under complicated background |
| CN104103080A (en) * | 2014-07-02 | 2014-10-15 | 华中科技大学 | Method of small dim target detection under complicated background |
| CN104766100B (en) * | 2014-10-22 | 2018-05-18 | 中国人民解放军电子工程学院 | Infrared small target image background Forecasting Methodology and device based on machine learning |
| CN104766100A (en) * | 2014-10-22 | 2015-07-08 | 中国人民解放军电子工程学院 | Infrared small target image background predicting method and device based on machine learning |
| CN105139432A (en) * | 2015-08-05 | 2015-12-09 | 西安电子科技大学 | Gaussian model based infrared small weak target image simulation method |
| CN105139432B (en) * | 2015-08-05 | 2018-03-06 | 西安电子科技大学 | Infrared DIM-small Target Image emulation mode based on Gauss model |
| CN105046713A (en) * | 2015-08-12 | 2015-11-11 | 北京航空航天大学 | Morphology-based robot star point segmentation method and FPGA realization method |
| CN105654516B (en) * | 2016-02-18 | 2019-03-26 | 西北工业大学 | Satellite image based on target conspicuousness is to ground weak moving target detection method |
| CN105654516A (en) * | 2016-02-18 | 2016-06-08 | 西北工业大学 | Method for detecting small moving object on ground on basis of satellite image with target significance |
| CN106327522A (en) * | 2016-08-25 | 2017-01-11 | 上海航天控制技术研究所 | Infrared small target detection method based on multi-direction morphological filtering complex cloud background |
| CN107564018A (en) * | 2017-08-30 | 2018-01-09 | 北京航空航天大学 | It is a kind of to utilize the method for improving iterative algorithm extraction target image |
| CN107976685A (en) * | 2017-11-20 | 2018-05-01 | 北京航空航天大学 | A kind of infrared sensor indoor human body Target Tracking System based on Internet of Things |
| CN108805878A (en) * | 2018-05-22 | 2018-11-13 | 深圳腾视科技有限公司 | A kind of foreground object detection solution based on computer vision |
| CN109712169A (en) * | 2018-11-15 | 2019-05-03 | 上海卫星工程研究所 | Moving-target motion profile prediction technique and method for autonomous tracking based on EO-1 hyperion |
| CN109859250A (en) * | 2018-11-20 | 2019-06-07 | 北京悦图遥感科技发展有限公司 | A kind of outer video multi-target detection of aviation red and tracking and device |
| CN109859250B (en) * | 2018-11-20 | 2023-08-18 | 北京悦图遥感科技发展有限公司 | Aviation infrared video multi-target detection and tracking method and device |
| CN109741396A (en) * | 2018-12-12 | 2019-05-10 | 天津津航技术物理研究所 | A kind of extremely small and weak infrared target detection method |
| CN109781142A (en) * | 2019-01-14 | 2019-05-21 | 上海卫星工程研究所 | Angle measurement is tested the speed Integrated Navigation Semi-physical Simulation verifying system and its test method |
| CN109903305B (en) * | 2019-01-24 | 2023-02-21 | 天津国为信息技术有限公司 | Linear target drop point positioning method based on aerial three-dimensional positioning |
| CN109903305A (en) * | 2019-01-24 | 2019-06-18 | 天津国为信息技术有限公司 | Line style target impact point positioning method based on aerial three-dimensional localization |
| CN109872345A (en) * | 2019-02-27 | 2019-06-11 | 中国科学院光电技术研究所 | A kind of monotrack method under dark background |
| CN109872345B (en) * | 2019-02-27 | 2022-08-26 | 中国科学院光电技术研究所 | Single target tracking method under dark background |
| CN109978851B (en) * | 2019-03-22 | 2021-01-15 | 北京航空航天大学 | Method for detecting and tracking small and medium moving target in air by using infrared video |
| CN109978851A (en) * | 2019-03-22 | 2019-07-05 | 北京航空航天大学 | A kind of aerial weak moving target detection tracking of infrared video |
| CN110502968B (en) * | 2019-07-01 | 2022-03-25 | 西安理工大学 | Method for detecting infrared small and weak moving target based on track point space-time consistency |
| CN110502968A (en) * | 2019-07-01 | 2019-11-26 | 西安理工大学 | The detection method of infrared small dim moving target based on tracing point space-time consistency |
| CN110728635A (en) * | 2019-09-10 | 2020-01-24 | 中国科学院上海技术物理研究所 | Contrast enhancement method for dim and weak target |
| CN110728635B (en) * | 2019-09-10 | 2023-07-07 | 中国科学院上海技术物理研究所 | Contrast enhancement method for dark and weak target |
| CN111583283A (en) * | 2020-05-20 | 2020-08-25 | 北京字节跳动网络技术有限公司 | Image segmentation method, image segmentation device, electronic device, and medium |
| CN112183221A (en) * | 2020-09-04 | 2021-01-05 | 北京科技大学 | Semantic-based dynamic object self-adaptive trajectory prediction method |
| CN112183221B (en) * | 2020-09-04 | 2024-05-03 | 北京科技大学 | Semantic-based dynamic object self-adaptive track prediction method |
| CN117095029A (en) * | 2023-08-22 | 2023-11-21 | 中国科学院空天信息创新研究院 | Method and device for detecting small target in air flight |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101567087B (en) | 2012-05-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101567087B (en) | Method for detecting and tracking small and weak target of infrared sequence image under complex sky background | |
| CN110517288B (en) | Real-time target detection tracking method based on panoramic multi-path 4k video images | |
| CN103077539B (en) | Motion target tracking method under a kind of complex background and obstruction conditions | |
| CN104835178A (en) | Low SNR(Signal to Noise Ratio) motion small target tracking and identification method | |
| CN106469313B (en) | Weak and small target detection method for pipe diameter self-adaptive time-space domain filtering | |
| CN110728697A (en) | Infrared dim target detection tracking method based on convolutional neural network | |
| CN107833239B (en) | Optimization matching target tracking method based on weighting model constraint | |
| CN107944354B (en) | Vehicle detection method based on deep learning | |
| CN113702979B (en) | Cross-region target tracking track segment space-time splicing method | |
| CN105913455A (en) | Local image enhancement-based object tracking method | |
| CN110400294B (en) | Infrared target detection system and detection method | |
| Dehghani et al. | Small target detection and tracking based on the background elimination and Kalman filter | |
| CN113763427B (en) | Multi-target tracking method based on coarse-to-fine shielding processing | |
| CN106327522A (en) | Infrared small target detection method based on multi-direction morphological filtering complex cloud background | |
| CN110349184B (en) | Multi-pedestrian tracking method based on iterative filtering and observation discrimination | |
| CN109448024B (en) | Visual tracking method and system for constructing constraint correlation filter by using depth data | |
| CN108038872B (en) | Dynamic and static target detection and real-time compressed sensing tracking research method | |
| Xu et al. | Moving target tracking based on adaptive background subtraction and improved camshift algorithm | |
| Bai et al. | Infrared small target detection and tracking under the conditions of dim target intensity and clutter background | |
| Sun et al. | Vehicle Flow Statistics System in Video Surveillance based on CAMSHIFT and Kalman Filter | |
| CN110660028B (en) | Small target detection method based on joint edge filtering morphology | |
| CN110211124B (en) | Infrared imaging frozen lake detection method based on MobileNet V2 | |
| Ganesan et al. | A Hybrid Face Image Contrast Enhancement Technique for Improved Face Recognition Accuracy. | |
| Wang et al. | Multiple feature fusion for tracking of moving objects in video surveillance | |
| Lee et al. | Hybrid real-time tracking of non-rigid objects under occlusions |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120523 Termination date: 20150525 |
|
| EXPY | Termination of patent right or utility model |