CN111311639B - Multi-search-space fast-moving self-adaptive update interval tracking method - Google Patents
Multi-search-space fast-moving self-adaptive update interval tracking method Download PDFInfo
- Publication number
- CN111311639B CN111311639B CN202010100756.9A CN202010100756A CN111311639B CN 111311639 B CN111311639 B CN 111311639B CN 202010100756 A CN202010100756 A CN 202010100756A CN 111311639 B CN111311639 B CN 111311639B
- Authority
- CN
- China
- Prior art keywords
- target
- frame
- tracking
- pos
- fast moving
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a tracking method of a fast moving self-adaptive update interval of multiple search spaces, which comprises the following steps: initializing a tracking filter template, target speed, acceleration and the like; calculating n approximate probability positions of the target by adopting multiple search spaces, and selecting one of the n approximate probability positions as an optimal target position; and judging whether the target is a fast moving target or not according to the target displacement amount, and respectively setting an updating interval strategy aiming at the fast moving target and a non-fast moving target. The invention provides a multi-search space strategy, and meanwhile, the updating interval of the tracking template is adjusted in a self-adaptive manner, so that the robustness of the tracking template is enhanced, and the problem of unstable performance of the tracker when the target moves fast is effectively solved.
Description
Technical Field
The invention relates to the field of computer vision, in particular to a tracking method for a fast moving self-adaptive update interval of multiple search spaces.
Background
Visual target tracking technology is one of the key research fields in the academic and industrial fields at present. Target tracking refers to a process of selecting a tracking target in a section of video or image sequence, searching a target position in a subsequent video or image sequence and recording target motion information. Currently, research achievements in the field of target tracking have been widely applied to the fields of intelligent video monitoring, human-computer interaction, intelligent navigation, three-dimensional reconstruction and the like. However, there are still many difficulties in the field of target tracking, in which the fast movement of the target is an important factor affecting the tracking performance.
In recent years, with intensive research on computer vision by many scholars, rapid development of the field of target tracking is promoted. A Minimum Output Sum of Squared Error filter (MOSSE) provided by Bolme et al firstly applies a correlation filter to a tracking field, greatly reduces the calculated amount and improves the tracking speed of a target, so that a target tracking algorithm based on the correlation filter gradually becomes the mainstream research direction of the target tracking field. Martin et al propose an effective Convolution operator Tracking algorithm (ECO), greatly improve the robustness and accuracy of target Tracking through factorized Convolution operation, compressed generation sample space operation and model washing and updating modes, and effectively reduce the calculated amount in the Tracking process. In the past, research on the fast movement of the target is less, and the fast movement of the target is one of important factors influencing the tracking performance of the target. Zhang superman et al propose a real-time tracking algorithm for a fast moving target under a complex background, and realize the tracking of the moving target through a background subtraction method of a Markov field model. The Sheng Liu et al provides a real-time fast moving target tracking algorithm for videos shot by an unmanned aerial vehicle, and fast movement and movement blurring are processed by analyzing optical flow information and adopting degradation distribution, so that tracking of fast moving targets can be achieved to a certain extent, but the tracking robustness is poor when the characteristics of the targets are changed greatly. Therefore, the influence of the target fast moving on the target tracking performance is still a problem to be solved urgently.
In order to enhance the tracking performance of the target under the condition of quick movement, the invention provides a multi-search space strategy aiming at the quick movement, predicts the target position by utilizing the speed, the acceleration and the like of the target, selects the optimal target position and improves the accuracy of target tracking; and the updating interval of the template is adaptively adjusted according to the predicted displacement, so that the possibility of template drift is reduced, and the target tracking performance is improved.
Disclosure of Invention
The invention relates to a tracking method for target rapid movement, aiming at solving the problem of poor tracking performance when a target rapidly moves.
The technical scheme of the invention is as follows:
a tracking method of fast moving adaptive update intervals of multiple search spaces comprises the following steps:
reading a video sequence to obtain a tracking target;
initializing a tracking template, target speed (speed and direction) and acceleration;
when the frame number t is less than or equal to 3, the center of the target frame tracked by the t-1 th frame is taken as the center coordinate, and the target center position pos is searched in a space with 4 times of the size (h multiplied by w) of the target frame t Then, step 6 is executed; when frame number t>3, select n search spaces to calculate pos 1 ,pos 2 ,…,pos n A total of n possible target center positions;
predicting the target center position p of the current frame according to the target speed and acceleration in the t-1 frame t And the total predicted displacement s sum ；
From pos 1 ,pos 2 ,…,pos n Selecting one of the positions and p t The position with small distance is used as the optimal central position of the target;
calculating the diagonal length d of the target frame in the last frame, i.e. the t-1 frame xy If, if
Judging the target to be a fast moving target, otherwise, judging the target to be a non-fast moving target;
setting different updating intervals N of the tracking template according to whether the target is in a fast moving state or not;
and updating the tracking template in time according to the interval N of the tracking template.
Repeating the steps 3 to 8 when t is t + 1; and completing target tracking until the last frame.
In step (3), when the number of frames t>When 3, the center of the target frame tracked by the t-1 th frame is taken as a central coordinate and is respectively positioned at m i Approximate probability positions of the target are searched in n spaces with the target frame size (h × w) multiplied by (i ═ 1,2, …, n), and pos is obtained 1 ,pos 2 ,…,pos n N possible target positions, where m i Is a real number greater than 1, pos i Is at m i The position in the range of multiple target frame size (h × w) where the correlation response value is maximum. In a conventional target tracking algorithm, a target search area is generally a rectangular area bounded by 2 to 5 times the size of a target, with the center coordinate of the target in the previous frame (t-1) being the center position. However, in the process of target tracking, if the set target search range is too large, excessive background information is introduced, which affects the accuracy of target tracking, and if the set tracking area is too small, when the target moves rapidly, the target easily exceeds the search area, which causes target tracking failure. The method provided by the invention searches in the ranges with different sizes respectively, and ensures that the target center position does not exceed the searching range.
In step (4), the speed can be decomposed into a lateral speed and a longitudinal speed, and the acceleration can be decomposed into a lateral acceleration and a longitudinal acceleration according to the physical meanings of the speed and the acceleration. The method is based on the transverse speed of the target in the t-1 frame
Longitudinal velocity
And lateral acceleration of the target
Longitudinal acceleration
Predicting the position of an object in a t-th frame
And the lateral predicted displacement s of the target in the t-1 to t frames x Longitudinal predicted displacement s y Total predicted displacement s sum 。
The calculation method of the transverse speed and the longitudinal speed comprises the following steps:
wherein the content of the first and second substances,
the coordinates of the target center position of the t-2 th frame and the t-1 th frame are respectively.
The calculation method of the transverse acceleration and the longitudinal acceleration comprises the following steps:
transverse predictive displacement s x Longitudinal predicted displacement s y And the predicted total predicted displacement s sum Respectively as follows:
it is known from the physical significance of velocity and acceleration that the change in velocity of the target is a continuous process that does not occur naturally. Therefore, the position and direction of the target in the current frame can be roughly predicted according to the speed and acceleration of the previous frame.
Predicted target position of t-th frame
Comprises the following steps:
in step (5), p t Is predicted from the velocity and acceleration of the t-1 frame, p t Although the target cannot be accurately located, the approximate position of the target is reflected. Therefore, the method considers the maximum response point, the distance p, obtained by searching in different ranges t The closer the confidence is greater. Therefore, the method calculates p t With pos 1 ,pos 2 ,…,pos n At pos of 1 ,pos 2 ,…,pos n In which a distance p is selected t The near is the center position where the target of the t-th frame is located. The calculation method comprises the following steps:
in the step (6), according to the diagonal length d of the target frame in the t-1 frame xy And the predicted displacement amount s sum Size determination target ofWhether in the fast moving state or the non-fast moving state. Predicted displacement s of target sum Greater than the target frame diagonal length d in the t-1 frame xy 1/2, the present invention considers the object to be in a fast moving state at this time, whereas the present invention considers the object to be in a slow moving state. The tracked targets in different videos are different in size, and even if the moving speed of the large target is slow, the displacement of the target center point is large. Even if a small target moves at a high speed, the displacement of the target center point is small. Therefore, when determining whether the target is in the fast moving state, if the threshold is set to a fixed value, it is not possible to accurately determine whether the target is in the fast moving state. Therefore, the present invention depends on the displacement of the center point of the target and the length d of the diagonal line of the target xy And comparing to determine whether the target is in a fast moving state.
In step (7), the invention determines the size of the update interval of the tracking template according to whether the tracked target is in a fast moving state. The update interval of the template is set to a small value u when the target is in a fast-moving state 1 Otherwise, the updating interval of the tracking template is set to a larger value u 2 . In the conventional algorithm, the tracking template is usually updated frame by frame or once every few frames. When the target is in a fast moving state, the possibility of target feature change is greatly increased, and if the tracking template is not updated timely at the moment, template drift is easily caused, and finally tracking failure is caused. When the target is in a slow motion state, if the update interval of the tracking template is too small, the situation of overfitting of the template caused by the redundancy of the tracking template is easy to occur. Therefore, there is a certain disadvantage in updating the template frame by frame or updating the template every fixed frame number. The invention adaptively adjusts the update interval of the tracking template of the target according to the state of the moving speed of the target, and effectively prevents the template from drifting.
Compared with the prior art, the method provided by the invention has the advantages that:
(1) the method has the advantages that the approximate position of the target is found under the multiple search spaces, and the tracking target which moves rapidly is ensured to be in the search range.
(2) Predicting the target position according to the speed and the acceleration to obtain a predicted point, calculating the Euclidean distance between each position in the multi-search space and the predicted point, and selecting the position with the minimum distance as the central position of the target. The method effectively improves the accuracy of target tracking.
(3) And judging whether the target is in a fast moving state or not according to the predicted target displacement and the diagonal length of the target in the previous frame, and adaptively adjusting the updating interval of the tracking template, thereby effectively improving the robustness of target tracking.
Drawings
FIG. 1 is a schematic flow chart of the algorithm of the present invention;
FIG. 2 is a schematic diagram of the success rate of the algorithm of the present invention and the ECO-HC algorithm in fast moving video;
FIG. 3 is a diagram showing the success rate of the algorithm of the present invention and the ECO-HC algorithm in the VOT2016 database.
The implementation, functional features and advantages of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It is to be understood that the described embodiments are merely a few embodiments of the invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the descriptions relating to "first", "second", etc. in the present invention are for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In addition, technical solutions between various embodiments may be combined with each other, but must be realized by a person skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination should not be considered to exist, and is not within the protection scope of the present invention.
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Referring to fig. 1, the example provided by the present invention uses the ECO-HC tracking algorithm as a reference filter, and the specific implementation manner is as follows.
(1) And reading the video sequence to obtain a tracking target.
(2) Initializing lateral velocity of tracked object
(pixel/frame), longitudinal velocity
(pixel/frame), lateral acceleration
(Pixel) 2 Frame), longitudinal acceleration
(Pixel) 2 Frame) and the angle theta between the motion speed and the horizontal direction is 0 0 。
(3) When the frame number t is less than or equal to 3, the center of the tracking target frame of the t-1 th frame is taken as a central coordinate, and the position pos of the target is searched in a space which is 4 times the size (h multiplied by w) of the target frame t (ii) a When the number of frames t>When 3, the center of the target frame tracked by the t-1 th frame is taken as a central coordinate and is respectively positioned at m i Approximate probability positions of the target are searched in n spaces with the target frame size (h × w) multiplied by (i ═ 1,2, …, n), and pos is obtained 1 ,pos 2 ,…,pos n N possible target positions, where m i Real numbers greater than 1. In the experiment of the invention, n is 2, m 1 ＝4，m 2 ＝5。
(4) According to the transverse speed of the target in the t-1 frame
Longitudinal velocity
Velocity direction theta and lateral acceleration
Longitudinal acceleration
Predicting the lateral displacement s of the current frame x Longitudinal displacement s y Target center position
And the total predicted displacement s sum 。
(5) Calculating a predicted position p t With pos 1 ,pos 2 ,…,pos n The calculation formula is as follows:
selection of d i Pos corresponding to the minimum value of (i-1, 2, …, n) i And taking the position as the central position of the t frame target. In the experiments of the present invention, n is 2.
(6) The length d of the diagonal of the target frame in the previous frame, i.e. the t-1 frame, is calculated xy If, if
Judging the target to be a fast moving target, otherwise, judging the target to be a non-fast moving target;
(7) setting different updating intervals N of the tracking template according to whether the target is in a fast moving state;
(8) and updating the tracking template in time according to the interval N of the tracking template.
(9) Repeating the steps 3 to 8 when t is t + 1; and completing target tracking until the last frame.
In order to verify the tracking performance of the invention, all videos in the data set VOT2016 and 8 videos (ball2, basketball, birds1, butterfly, gynmantics 4, iceskater1, fernando and handball2) which are rapidly affected by a target in the VOT2016 are selected to be tested respectively, and the tracking performance is represented by Robustness (Robustness), Accuracy (Accuracy), overlapping rate (overlay) and error rate (failurs), Average overlapping Expectation (EAO) and Area under the curve (AUC). The results of the experiment are shown in tables 1 and 2 and fig. 2 and 3.
TABLE 1 analysis of the Performance of the inventive method and ECO-HC experiments on target fast moving data set
| EAO | Robustness | Accuracy | Overlap | Failures | AUC | |
| Our | 0.36 | 2.33 | 0.43 | 0.50 | 2.83 | 0.49 |
| ECO-HC | 0.26 | 3.17 | 0.42 | 0.48 | 4.84 | 0.35 |
As can be seen from Table 2, compared with ECO-HC, the average overlap Expectation (EAO) of the algorithm provided by the invention is improved by 10% when the algorithm is used for dealing with the target fast movement, the robustness of the algorithm is improved by 0.84 compared with the ECO-HC, the accuracy of the algorithm is improved by 1% compared with the ECO-HC, the overlap rate of the algorithm is improved by 2% compared with the ECO-HC, the error rate of the algorithm is reduced by 2.01 compared with the ECO-HC, and the area under the curve (AUC) of the algorithm is improved by 14% compared with the ECO-HC.
TABLE 2 Experimental results of the algorithm of the present invention and ECO-HC in VOT2016
| EAO | Robustness | Accuracy | Overlap | Failures | AUC | |
| Our | 0.33 | 17.83 | 0.56 | 0.55 | 20.24 | 0.42 |
| ECO-HC | 0.32 | 19.00 | 0.54 | 0.53 | 21.40 | 0.40 |
As can be seen from Table 2, compared with the average overlap Expectation (EAO) of ECO-HC, the algorithm provided by the invention has improved 1 percentage point, the Robustness (Robustness) of the algorithm provided by the invention has improved 1.17 compared with ECO-HC, the Accuracy (Accuracy) of the algorithm provided by the invention has improved 2% compared with ECO-HC, the overlap rate of the algorithm provided by the invention has improved 2% compared with ECO-HC, the error rate (failurs) of the algorithm provided by the invention has reduced 1.16 compared with ECO-HC, and the area under the curve (AUC) of the algorithm provided by the invention has improved 2 percentage points compared with ECO-HC.
As can be seen from FIG. 2, the algorithm success rate curve proposed by the present invention is significantly better than that of ECO-HC under the target fast moving data set. As can be seen from fig. 3, the success rate curve of the algorithm proposed by the present invention is significantly better than that of ECO-HC under the VOT2016 data set.
In conclusion, the fast moving self-adaptive update interval tracking method with multiple search spaces provided by the invention effectively improves the accuracy and robustness of target tracking and has extremely high popularization value.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (7)
1. A tracking method for fast moving adaptive update intervals of multiple search spaces is characterized by comprising the following steps:
step (1): reading a video frame sequence to obtain an initial frame image;
step (2): initializing a tracking template and the speed and the acceleration of a tracking target according to the target information of the initial frame;
and (3): when the frame number t is less than or equal to 3, the center of the target frame tracked by the t-1 th frame is taken as a central coordinate, and the target central position pos is searched in a space which is 4 times the size h multiplied by w of the target frame t Then, step (6) is executed; when the number of frames t>At time 3, with the center of the tracking target frame of the t-1 th frame as a central coordinate, searching the approximate probability position of the target in n spaces with mi being 1,2, … and n times of the size h multiplied by w of the target frame, and obtaining pos 1 ,pos 2 ,…,pos n N possible target positions, where m i Is a real number greater than 1, pos i Is at m i The position with the maximum correlation response value in the range h multiplied by w times the size of the target frame;
and (4): predicting the target center position p of the current frame according to the target speed and acceleration in the t-1 frame t ；
And (5): from pos 1 ,pos 2 ,…,pos n Selecting one of the positions and p t The position with small distance is used as the optimal central position of the target;
and (6): calculating the diagonal length d of the target frame in the t-1 frame xy And the total predicted displacement s sum If at all
Judging the target to be a fast moving target, otherwise, judging the target to be a non-fast moving target;
and (7): respectively setting a tracking template updating interval N aiming at a fast moving target and a non-fast moving target;
and (8): updating the template in time according to the updating interval N;
and (9): and (5) repeating the steps 3 to 8 until the tracking is finished, wherein t is t + 1.
2. The method for tracking fast moving adaptive update interval of multiple search spaces according to claim 1, wherein the step (2) is implemented as follows: initializing a tracking template F1 by using a correlation filter algorithm according to the 1 st frame target, and initializing the transverse speed of the tracking target
(pixel/frame), longitudinal velocity
(pixel/frame), lateral acceleration
(Pixel) 2 Frame), longitudinal acceleration
(Pixel) 2 Frame) and the angle theta between the motion speed and the horizontal direction is 0 0 。
3. The method for tracking fast moving adaptive update interval of multiple search spaces according to claim 1, wherein the step (3) is implemented as follows: when the frame number t is less than or equal to 3, the center of the tracking target frame of the t-1 th frame is taken as a central coordinate, and the position pos of the target is searched in a space with 4 times of the size of the target frame h multiplied by w t (ii) a When the number of frames t>When 3, the center of the target frame tracked by the t-1 th frame is taken as a central coordinate and is respectively positioned at m i Where i is 1,2, …, the target frame size n times is n spaces h × w, the approximate probability position of the target is searched, and pos is obtained 1 ,pos 2 ,…,pos n N possible target positions, where m i Real numbers greater than 1.
4. The method for tracking fast moving adaptive update interval of multiple search spaces according to claim 1, wherein the step (4) is implemented as follows: is provided with
The coordinates of the center position of the target in the t-2 th frame and the t-1 th frame respectively, the speed v of the target t-1 Is calculated by the formula
In the formula (I), the compound is shown in the specification,
as transverse velocity
Is the longitudinal velocity
Where Δ t has a value of 1, θ is the angle between the speed of motion and the horizontal and ranges from
It is to be noted that
Positive or negative of (b) indicates the direction of movement of the objectThe direction of the solution is as follows; is provided with
The running speeds of the target at the t-2 th frame and the t-1 th frame respectively, the acceleration a of the target t-1 Is calculated by the formula
In the formula (I), the compound is shown in the specification,
in order to be the lateral acceleration,
for longitudinal acceleration, where Δ t has a value of 1,
positive and negative of (d) indicate the direction of acceleration; predicting the target position of the current t frame according to the speed and acceleration of the target in the t-1 frame
And the total predicted displacement s sum The calculation formula is
Here, the value of Δ t is generally 1.
5. The method for tracking fast moving adaptive update interval of multiple search spaces according to claim 1, wherein said step (5) is performed according to the predicted position p t As evaluation criteria, pos was calculated 1 ,pos 2 ,…,pos n And p t The calculation formula is as follows:
selection of d i Where i is 1,2, …, the minimum value of n corresponds to pos i As the optimum position.
6. The method for tracking fast moving adaptive update interval in multiple search spaces according to claim 1, wherein said step (7) sets the update interval N, and sets a small update interval u for fast moving objects 1 Because the target state changes greatly when the target moves rapidly, the target information can be updated into the template as soon as possible at a small updating interval; while a large update interval u is set for non-fast moving objects 2 。
7. The method for tracking fast moving adaptive update interval of multiple search spaces according to claim 1, wherein said step (8) updates the tracking filter template every N frames according to the update interval N; the update interval N of the tracking filter template is a dynamic value and is determined according to the moving speed of the target.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2019114098667 | 2019-12-31 | ||
| CN201911409866 | 2019-12-31 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111311639A CN111311639A (en) | 2020-06-19 |
| CN111311639B true CN111311639B (en) | 2022-08-26 |
Family
ID=71148386
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010100756.9A Active CN111311639B (en) | 2019-12-31 | 2020-02-19 | Multi-search-space fast-moving self-adaptive update interval tracking method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111311639B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103617631A (en) * | 2013-11-11 | 2014-03-05 | 山东神戎电子股份有限公司 | Tracking method based on center detection |
| CN106097388A (en) * | 2016-06-07 | 2016-11-09 | 大连理工大学 | In video frequency object tracking, target prodiction, searching scope adaptive adjust and the method for Dual Matching fusion |
| CN107066990A (en) * | 2017-05-04 | 2017-08-18 | 厦门美图之家科技有限公司 | A kind of method for tracking target and mobile device |
| CN107301657A (en) * | 2017-06-12 | 2017-10-27 | 西安交通大学 | A kind of video target tracking method for considering target movable information |
| CN108182447A (en) * | 2017-12-14 | 2018-06-19 | 南京航空航天大学 | A kind of adaptive particle filter method for tracking target based on deep learning |
| CN110516528A (en) * | 2019-07-08 | 2019-11-29 | 杭州电子科技大学 | A kind of moving-target detection and tracking method based under movement background |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6509027B2 (en) * | 2015-05-12 | 2019-05-08 | キヤノン株式会社 | Object tracking device, optical apparatus, imaging device, control method of object tracking device, program |
-
2020
- 2020-02-19 CN CN202010100756.9A patent/CN111311639B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103617631A (en) * | 2013-11-11 | 2014-03-05 | 山东神戎电子股份有限公司 | Tracking method based on center detection |
| CN106097388A (en) * | 2016-06-07 | 2016-11-09 | 大连理工大学 | In video frequency object tracking, target prodiction, searching scope adaptive adjust and the method for Dual Matching fusion |
| CN107066990A (en) * | 2017-05-04 | 2017-08-18 | 厦门美图之家科技有限公司 | A kind of method for tracking target and mobile device |
| CN107301657A (en) * | 2017-06-12 | 2017-10-27 | 西安交通大学 | A kind of video target tracking method for considering target movable information |
| CN108182447A (en) * | 2017-12-14 | 2018-06-19 | 南京航空航天大学 | A kind of adaptive particle filter method for tracking target based on deep learning |
| CN110516528A (en) * | 2019-07-08 | 2019-11-29 | 杭州电子科技大学 | A kind of moving-target detection and tracking method based under movement background |
Non-Patent Citations (2)
| Title |
|---|
| High-speed Tracking with Kernelized Correlation Filters;Henriques,JF.Caseiro;《IEEE Transactions on Pattern Analysis & Machine Intelligence》;20151231;全文 * |
| 自适应模板更新和目标重定位的相关滤波器跟踪;刘教民，郭剑威，师硕;《光学精密工程》;20180831;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111311639A (en) | 2020-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109636829B (en) | Multi-target tracking method based on semantic information and scene information | |
| CN110751096B (en) | Multi-target tracking method based on KCF track confidence | |
| CN110796676A (en) | Target tracking method combining high-confidence updating strategy with SVM (support vector machine) re-detection technology | |
| CN112052802B (en) | Machine vision-based front vehicle behavior recognition method | |
| CN114638855A (en) | Multi-target tracking method, equipment and medium | |
| CN107844739B (en) | Robust target tracking method based on self-adaptive simultaneous sparse representation | |
| CN111862155A (en) | Unmanned aerial vehicle single vision target tracking method aiming at target shielding | |
| KR20210027778A (en) | Apparatus and method for analyzing abnormal behavior through object detection and tracking | |
| CN110111370B (en) | Visual object tracking method based on TLD and depth multi-scale space-time features | |
| CN111639570B (en) | Online multi-target tracking method based on motion model and single-target clue | |
| Zhang et al. | A background-aware correlation filter with adaptive saliency-aware regularization for visual tracking | |
| CN111429485B (en) | Cross-modal filtering tracking method based on self-adaptive regularization and high-reliability updating | |
| CN113537077A (en) | Label multi-Bernoulli video multi-target tracking method based on feature pool optimization | |
| CN110660084A (en) | Multi-target tracking method and device | |
| CN111311639B (en) | Multi-search-space fast-moving self-adaptive update interval tracking method | |
| CN111768427B (en) | Multi-moving-object tracking method, device and storage medium | |
| CN111986231A (en) | Multi-target tracking method and system | |
| CN115861386A (en) | Unmanned aerial vehicle multi-target tracking method and device through divide-and-conquer association | |
| CN107194947B (en) | Target tracking method with self-adaptive self-correction function | |
| CN115272393A (en) | Video stream target tracking method and device for unmanned aerial vehicle and storage medium | |
| CN115761693A (en) | Method for detecting vehicle location mark points and tracking and positioning vehicles based on panoramic image | |
| CN117170354A (en) | Wheel type robot positioning method and system | |
| CN112037272B (en) | Target tracking method for rapid regional search based on correlation filtering | |
| CN112465865A (en) | Multi-target tracking method based on background modeling and IoU matching | |
| CN114973195A (en) | Vehicle tracking method, device and system based on multi-information fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |