CN111583306A - Anti-occlusion visual target tracking method - Google Patents
Anti-occlusion visual target tracking method Download PDFInfo
- Publication number
- CN111583306A CN111583306A CN202010398545.8A CN202010398545A CN111583306A CN 111583306 A CN111583306 A CN 111583306A CN 202010398545 A CN202010398545 A CN 202010398545A CN 111583306 A CN111583306 A CN 111583306A
- Authority
- CN
- China
- Prior art keywords
- target
- frame
- tracking
- filter
- cnn
- 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.)
- Pending
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
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
-
- 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
- G06T2207/10024—Color image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an anti-occlusion visual target tracking method, and belongs to the field of pattern recognition and computer vision. According to the method, the area of the target is found out in the rest image frames according to the position parameters of an input video and the initial frame of the target to be tracked. And judging the tracking quality by combining the peak side lobe ratio according to an APSR strategy to determine whether to update a related filter, and solving the problem of target tracking failure under the conditions of shielding or deformation of a tracked target in a video and the like. The invention solves the problem of invalid updating of the filter under the condition of target shielding or deformation, thereby improving the tracking precision in actual tracking and having important significance on a video target tracking algorithm and subsequent behavior analysis of a specific target.
Description
Technical Field
The invention belongs to the field of pattern recognition and computer vision, and relates to an anti-blocking visual target tracking method which can be applied to the fields of intelligent video monitoring, intelligent driving, unmanned aerial vehicle monitoring and the like.
Background
Video motion target tracking is one of the popular technologies in the current computer field, target behaviors in videos can be accurately tracked and positioned, the target tracking theory is more and more perfect along with the continuous updating of an algorithm, and the application field relates to intelligent video monitoring, unmanned aerial vehicle reconnaissance, intelligent driving and the like. In brief, the target tracking means that the initial position of the target is given in the first frame, and the position information of the target in each subsequent frame of image is calculated by using a tracking algorithm. Theoretically, target tracking can be performed in real time, but in practical application, due to factors such as illumination, shielding, scale change and the like, a target is easily lost.
In general, a target tracking algorithm can be classified into a generator method and a discriminant method from the viewpoint of constructing a target model. And (3) carrying out feature extraction and model construction on the target by a generative method, and finding a region similar to the model in the next frame, namely the prediction region of the target. The discriminant method attributes the tracking problem to a binary classification problem, and mainly studies how to distinguish the target from the background.
Compared with the two methods, the discriminant method can be more suitable for complex problems such as background change and the like. The discriminant method is continuously improved in recent years, great breakthrough is made in the technical aspect, and researchers continuously improve the algorithm from the aspects of characteristics, scales and the like, so that the target tracking is more suitable for complex and changeable environments.
Due to the continuous change of target characteristics caused by the change of target and environment information in the target tracking process and the requirements of target tracking on tracking speed and precision, the target tracking has the following main difficulties:
1) the appearance of the object changes. The appearance of the target changes due to the movement of the object and the non-rigid deformation (such as jumping and walking of people), or the appearance of the target changes due to the change of the shooting angle.
2) And (4) dimension change. The size of the area occupied by the target in the image changes due to factors such as the shooting distance.
3) The environment changes. The imaging characteristics of the target image and the like are changed due to the change of the shooting environment (such as illumination, weather and the like).
4) The object moves rapidly. The rapid movement of the target causes abrupt change of the coordinate position in the image, and the speed and the precision of target searching are influenced.
5) And (5) blocking the target and getting out of the visual field. The method is characterized by comprising the following steps of partially or completely losing features due to the fact that a target is shielded by other objects in shooting, and failing to track due to the fact that the target jumps out of a visual field to track again in shooting.
6) The imaging effect. Due to the fact that the infrared camera is low in resolution, the target resolution is low, the edge and target features are not obvious, sometimes, the difference between a target and an environment is small, focusing problems of task equipment and the like can cause difficulty in feature extraction during target tracking, and the like.
The factors have great influence on target feature extraction and a target search strategy in target tracking, and in the actual tracking process, the influence caused by the factors is accurately and timely processed, so that the accuracy and robustness of target tracking can be ensured.
Disclosure of Invention
In view of the above, the present invention provides an anti-occlusion visual target tracking method, which solves the problem of poor accuracy in target tracking in the prior art. The method judges the tracking quality by using an APSR strategy, solves the problem of invalid updating of a filter under the condition that a target is shielded or deformed, and accordingly improves the tracking precision in actual tracking.
In order to achieve the purpose, the invention provides the following technical scheme:
an anti-occlusion visual target tracking method specifically comprises the following steps:
s1: reading a video sequence to be tracked, determining the position information of a target to be tracked according to the target to be tracked of a given initial frame, extracting the characteristics of CN, CNN and Hog, and calculating a weight graph G of a t framet;
S2: training a relevant filter f by using the extracted features of the target CN, the CNN and the Hog;
s3: inputting a t +1 frame image, searching a target position in the t +1 frame, extracting CN, CNN and Hog characteristics, and solving a filter f in the t frametCalculating a target response graph K of the t +1 framet+1;
S4: target response map K of t +1 th framet+1Judging the tracking quality by adopting an APSR strategy and combining a peak side lobe ratio to determine whether to update a related filter;
s5: and forming a target position frame by combining the target position information of all the frames, generating a video for calibrating a target area, and outputting the video, thereby facilitating subsequent analysis and research.
Further, in the step S1, the weight map G of the t-th frametComposed of a superposition of a spatially perceptual weight map Q and a target-likeness weight map R, i.e. Gt=Q+R;
Calculating a spatial perception weight map Q, specifically comprising: the numerical value of the spatial perception weight of any pixel pi is recorded as Q (pi), Q (pi) of all pixel points of the whole target frame is calculated, and a final Q is generated;
calculating a target similarity weight graph R, which specifically comprises the following steps: knowing the image of the t-th frame as an initial frame, constructing a required color histogram given the position and size of the targetAndthe expression is as follows:
wherein the content of the first and second substances,which represents the target of the t-th frame,representing the background color histogram of the t-th frame, gamma is a fixed update rate,andtarget and background color histograms for frame 1 through frame t-1, respectively, based onThe target similarity weight map R based on the color histogram is obtained by:
wherein the content of the first and second substances,andrespectively representing the proportion of the pixel value size of the target area and the background area of the t-th frame to the total pixel value of the whole given search area.
Further, in step S2, the correlation filter f is trained by using the CN, CNN and Hog features of the target region in the t-th frametThe actual optimization function is:
wherein x isk、ykTo satisfy the labels of the gaussian distribution, ω is the constraint of the filter; f. oft=ft⊙GtWhere ⊙ represents the dot product, and at the minimum (f), training the correlation filter f needed to obtain the t-th framet。
Further, the step S3 specifically includes: reading the image of the t +1 th frame, finding a target position in a given search area of the t +1 th frame, performing center cutting on the target position of the previous frame to obtain a search area, extracting CN, CNN and Hog characteristics of the search area, and determining the characteristics as zt+1In combination with ftTo calculate the final target response map K of the t +1 th framet+1The expression is:
Kt+1=F-1(f︿t⊙z︿t+1)
wherein f is︿tAnd z︿t+1Represents a pair of ftAnd zt+1Performing Fourier transform, F-1Representing an inverse fourier transform;
according to the target response chart Kt+1To obtain the t +1 th frameThe actual target position of.
Further, the step S4 specifically includes: target response map K for frame t +1t+1Tracking quality according to an APSR strategy and a peak-to-side lobe ratio; wherein APSR is defined as follows:
wherein, KmaxAnd KminAre each Kt+1Maximum and minimum values of, the region around the peakIs μ and the standard deviation is1And σ1(ii) a Wherein, Kt+1Is notAll ranges are noted asτ is a given parameter (τ ≈ 1.0), w and h are each Kt+1Horizontal and vertical coordinates of middle pixel, Kw,hIs Kt+1The corresponding value of the coordinate (w, h) of (a), mean is the averaging function;
and calculating the value of the APSR, representing the tracking quality, and influencing whether the correlation filter of the t-th frame is updated or not by combining the peak-to-side lobe ratio.
The invention has the beneficial effects that:
(1) the invention can effectively process the tracking of the target under complex scenes such as shielding, deformation and the like. In the visual target tracking under the actual scene, the method utilizes a spatial punishment mechanism to train a relevant filter, can obtain a filter capable of effectively locking correct target pixels, and also weakens the influence of background pixels to a certain extent. Therefore, the learned filter has certain memorability, when the tracked target disappears in the sight line for a short time, the tracker can confirm that the tracked target does not exist in the current search area any more, the training model (avoiding being polluted by background pixels) at the moment can be stopped from being updated, the tracker with the correct characteristic information of the target is reserved, and the target can still wait for reappearance of the target in the subsequent frame and be tracked.
(2) The tracking algorithm of the invention is less time consuming. The method has the advantages of high tracking speed, which is beneficial to the speed advantage of a related filtering algorithm, and because the method has simple and direct optimization process, an ideal filtering template can be quickly trained based on the idea of loop iteration.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.
Drawings
For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
FIG. 1 is an overall work flow diagram of the anti-occlusion target tracking method of the present invention;
FIG. 2 is a flowchart of the process of calculating the weight map for the t-th frame.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
Referring to fig. 1 to 2, fig. 1 is an overall work flow diagram of an anti-occlusion target tracking method provided by the present invention, as shown in fig. 1, the method specifically includes the following steps:
step 1: reading a video sequence to be tracked, determining the position information of a target to be tracked according to the target to be tracked of a given initial frame, extracting the characteristics of CN, CNN and Hog, and calculating a weight map of a t-th frame, as shown in FIG. 2, specifically comprising:
defining the weight map of the t-th frame as GtThe weight graph is formed by superposing a spatial perception weight graph Q and a target similarity weight graph R;
the features to be extracted include CN, CNN and Hog features, the spatial perception weight graph Q is obtained, and the weight value of each pixel point in the target frame is attenuated along with the distance from the target center. The numerical value of the spatial perception weight of any pixel pi is recorded as Q (pi), Q (pi) of all pixel points of the whole target frame is calculated, and a final Q is generated;
calculating a target similarity weight map R: knowing the image of the t-th frame as the initial frame, given the target position and size, we can construct the desired color histogramAndas follows:
wherein, among others,which represents the target of the t-th frame,representing the background color histogram of the t-th frame, gamma is a fixed update rate,andthe target and background color histograms for frame 1 to frame t-1, respectively, may be obtained from a color histogram based target similarity weight map R according to the following equation:
wherein the content of the first and second substances,andrespectively representing the proportion of the pixel value of the target area and the background area of the t-th frame to the total pixel value of the whole given search area;
the above steps obtain a spatial perception weight map Q and a target similar weight map R, and then the final weight map G of the t-th frametCalculated by the following formula:
Gt=Q+R
step 2: and training the correlation filter f by using the extracted target CN, CNN and Hog characteristics. The method specifically comprises the following steps: the CN, CNN and Hog characteristics of the target area of the t frame are extracted, and a correlation filter f can be trainedtThe actual optimization function is also as follows:
in the above formula, x is definedk,ykThe labels satisfy Gaussian distribution, and omega is the constraint of the filter; f. oft=ft⊙GtWherein ⊙ represents the dot product, and training the correlation filter f needed to obtain the t-th frame when (f) is minimumt。
And step 3: inputting a t +1 frame image, searching a target position in the t +1 frame, extracting CN, CNN and Hog characteristics, and solving a filter f in the t frametCalculating a t +1 frame response map Kt+1. The method specifically comprises the following steps:
reading the image of the t +1 th frame, finding a target position in a given search area of the t +1 th frame, performing center cutting on the target position of the previous frame to obtain a search area, extracting CN, CNN and Hog characteristics of the search area, and determining the characteristics as zt+1In combination with ftTo calculate the final response map K of the t +1 th framet+1:
Kt+1=F-1(f︿t⊙z︿t+1)
Wherein f is︿tAnd z︿t+1Namely, is to ftAnd zt+1Performing Fourier transform, F-1I.e. the inverse Fourier transform, Kt+1Is the target response map of the t +1 th frame;
according to the response chart Kt+1And obtaining the actual target position of the t +1 th frame.
And 4, step 4: target response map K of t +1 th framet+1And judging the tracking quality by adopting an APSR strategy and combining a peak-to-side lobe ratio, and determining whether to update a correlation filter. The method specifically comprises the following steps:
target response map K for the t +1 th frame calculated in step 3t+1The tracking quality can be known according to the APSR strategy and the peak-to-side lobe ratio, where APSR is defined as follows:
wherein, Kt+1Maximum value of (A) is Kmax,Kt+1Is KminNear peak regionIs μ1Near peak regionHas a standard deviation of1Wherein, K ist+1Is notAll ranges are noted asτ is a given parameter (τ ≈ 1.0), w and h are Kt+1Horizontal and vertical coordinates of middle pixel, Kw,hIs Kt+1The corresponding value of the coordinate (w, h) of (a), mean is the averaging function;
and calculating the value of the APSR, which can represent the tracking quality, and can influence whether the correlation filter of the t-th frame is updated or not by combining the peak-to-side lobe ratio.
And 5: and forming a target position frame by combining the target position information of all the frames, generating a video for calibrating a target area, and outputting the video, thereby facilitating subsequent analysis and research.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.
Claims (5)
1. An anti-occlusion visual target tracking method is characterized by specifically comprising the following steps:
s1: reading a video sequence to be tracked, determining the position information of a target to be tracked according to the target to be tracked of a given initial frame, extracting the characteristics of CN, CNN and Hog, and calculating a weight graph G of a t framet;
S2: training a relevant filter f by using the extracted features of the target CN, the CNN and the Hog;
s3: inputting a t +1 frame image, searching a target position in the t +1 frame, extracting CN, CNN and Hog characteristics, and solving a filter f in the t frametCalculating a target response graph K of the t +1 framet+1;
S4: target response map K of t +1 th framet+1Judging the tracking quality by adopting an APSR strategy and combining a peak side lobe ratio to determine whether to update a related filter;
s5: and combining the target position information of all the frames to form a target position frame and generating a video of a calibration target area.
2. The method for tracking an anti-occlusion visual target according to claim 1, wherein in the step S1, the weight map G of the t frametComposed of a superposition of a spatially perceptual weight map Q and a target-likeness weight map R, i.e. Gt=Q+R;
Calculating a spatial perception weight map Q, specifically comprising: the numerical value of the spatial perception weight of any pixel pi is recorded as Q (pi), Q (pi) of all pixel points of the whole target frame is calculated, and a final Q is generated;
calculating a target similarity weight graph R, which specifically comprises the following steps: knowing the image of the t-th frame as an initial frame, constructing a required color histogram given the position and size of the targetAndthe expression is as follows:
wherein the content of the first and second substances,which represents the target of the t-th frame,representing the background color histogram of the t-th frame, gamma is a fixed update rate,andrespectively obtaining target similarity weight graph R based on the color histogram according to the following formula from the 1 st frame to the t-1 st frame as well as the background color histogram:
3. The method for tracking an anti-occlusion visual target according to claim 2, wherein in the step S2, the correlation filter f is trained by using CN, CNN and Hog features of the target region of the t-th frametThe actual optimization function is:
wherein x isk、ykTo satisfy the labels of the gaussian distribution, ω is the constraint of the filter; f. oft=ft⊙GtWhere ⊙ represents the dot product, and at the minimum (f), training the correlation filter f needed to obtain the t-th framet。
4. The anti-occlusion visual target tracking method according to claim 3, wherein the step S3 specifically comprises: reading the image of the t +1 th frame, finding a target position in a given search area of the t +1 th frame, performing center cutting on the target position of the previous frame to obtain a search area, extracting CN, CNN and Hog characteristics of the search area, and determining the characteristics as zt+1Knot ofAnd ftTo calculate the final target response map K of the t +1 th framet+1The expression is:
Kt+1=F-1(f︿t⊙z︿t+1)
wherein f is︿tAnd z︿t+1Represents a pair of ftAnd zt+1Performing Fourier transform, F-1Representing an inverse fourier transform;
according to the target response chart Kt+1And obtaining the actual target position of the t +1 th frame.
5. The anti-occlusion visual target tracking method according to claim 4, wherein the step S4 specifically comprises: target response map K for frame t +1t+1Tracking quality according to an APSR strategy and a peak-to-side lobe ratio; wherein APSR is defined as follows:
wherein, KmaxAnd KminAre each Kt+1Maximum and minimum values of, the region around the peakIs μ and the standard deviation is1And σ1(ii) a Wherein, Kt+1Is notAll ranges are noted asτ is a given parameter, w and h are each Kt+1Horizontal and vertical coordinates of middle pixel, Kw,hIs Kt+1The corresponding value of the coordinate (w, h) of (a), mean is the averaging function;
and calculating the value of the APSR, representing the tracking quality, and influencing whether the correlation filter of the t-th frame is updated or not by combining the peak-to-side lobe ratio.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010398545.8A CN111583306A (en) | 2020-05-12 | 2020-05-12 | Anti-occlusion visual target tracking method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010398545.8A CN111583306A (en) | 2020-05-12 | 2020-05-12 | Anti-occlusion visual target tracking method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111583306A true CN111583306A (en) | 2020-08-25 |
Family
ID=72112264
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010398545.8A Pending CN111583306A (en) | 2020-05-12 | 2020-05-12 | Anti-occlusion visual target tracking method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111583306A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114359337A (en) * | 2021-12-07 | 2022-04-15 | 中国人民解放军国防科技大学 | RGBT visual target tracking method and device, electronic equipment and storage medium |
CN116563348A (en) * | 2023-07-06 | 2023-08-08 | 中国科学院国家空间科学中心 | Infrared weak small target multi-mode tracking method and system based on dual-feature template |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109785366A (en) * | 2019-01-21 | 2019-05-21 | 中国科学技术大学 | It is a kind of for the correlation filtering method for tracking target blocked |
CN110211157A (en) * | 2019-06-04 | 2019-09-06 | 重庆邮电大学 | A kind of target long time-tracking method based on correlation filtering |
CN110458862A (en) * | 2019-05-22 | 2019-11-15 | 西安邮电大学 | A kind of motion target tracking method blocked under background |
CN110599519A (en) * | 2019-08-27 | 2019-12-20 | 上海交通大学 | Anti-occlusion related filtering tracking method based on domain search strategy |
CN111091583A (en) * | 2019-11-22 | 2020-05-01 | 中国科学技术大学 | Long-term target tracking method |
-
2020
- 2020-05-12 CN CN202010398545.8A patent/CN111583306A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109785366A (en) * | 2019-01-21 | 2019-05-21 | 中国科学技术大学 | It is a kind of for the correlation filtering method for tracking target blocked |
CN110458862A (en) * | 2019-05-22 | 2019-11-15 | 西安邮电大学 | A kind of motion target tracking method blocked under background |
CN110211157A (en) * | 2019-06-04 | 2019-09-06 | 重庆邮电大学 | A kind of target long time-tracking method based on correlation filtering |
CN110599519A (en) * | 2019-08-27 | 2019-12-20 | 上海交通大学 | Anti-occlusion related filtering tracking method based on domain search strategy |
CN111091583A (en) * | 2019-11-22 | 2020-05-01 | 中国科学技术大学 | Long-term target tracking method |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114359337A (en) * | 2021-12-07 | 2022-04-15 | 中国人民解放军国防科技大学 | RGBT visual target tracking method and device, electronic equipment and storage medium |
CN116563348A (en) * | 2023-07-06 | 2023-08-08 | 中国科学院国家空间科学中心 | Infrared weak small target multi-mode tracking method and system based on dual-feature template |
CN116563348B (en) * | 2023-07-06 | 2023-11-14 | 中国科学院国家空间科学中心 | Infrared weak small target multi-mode tracking method and system based on dual-feature template |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Automatic laser profile recognition and fast tracking for structured light measurement using deep learning and template matching | |
CN109800689B (en) | Target tracking method based on space-time feature fusion learning | |
CN111797716B (en) | Single target tracking method based on Siamese network | |
CN111539273B (en) | Traffic video background modeling method and system | |
CN108010067B (en) | A kind of visual target tracking method based on combination determination strategy | |
CN106845374B (en) | Pedestrian detection method and detection device based on deep learning | |
CN106875424B (en) | A kind of urban environment driving vehicle Activity recognition method based on machine vision | |
CN108416266B (en) | Method for rapidly identifying video behaviors by extracting moving object through optical flow | |
CN111563915B (en) | KCF target tracking method integrating motion information detection and Radon transformation | |
CN110533695A (en) | A kind of trajectory predictions device and method based on DS evidence theory | |
JP7263216B2 (en) | Object Shape Regression Using Wasserstein Distance | |
CN111582349B (en) | Improved target tracking algorithm based on YOLOv3 and kernel correlation filtering | |
CN107833239B (en) | Optimization matching target tracking method based on weighting model constraint | |
CN110033472B (en) | Stable target tracking method in complex infrared ground environment | |
CN111724411B (en) | Multi-feature fusion tracking method based on opposite-impact algorithm | |
CN107424175B (en) | Target tracking method combined with space-time context information | |
CN106780560A (en) | A kind of feature based merges the bionic machine fish visual tracking method of particle filter | |
CN111583306A (en) | Anti-occlusion visual target tracking method | |
CN106127766B (en) | Method for tracking target based on Space Coupling relationship and historical models | |
CN107609571A (en) | A kind of adaptive target tracking method based on LARK features | |
CN115205903B (en) | Pedestrian re-recognition method based on identity migration generation countermeasure network | |
Liu et al. | Correlation filter with motion detection for robust tracking of shape-deformed targets | |
CN110517285B (en) | Large-scene minimum target tracking based on motion estimation ME-CNN network | |
CN117011381A (en) | Real-time surgical instrument pose estimation method and system based on deep learning and stereoscopic vision | |
CN117011342A (en) | Attention-enhanced space-time transducer vision single-target tracking method |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200825 |