CN113949812A - Electronic image stabilization method based on partitioned Kalman motion prediction - Google Patents
Electronic image stabilization method based on partitioned Kalman motion prediction Download PDFInfo
- Publication number
- CN113949812A CN113949812A CN202111229346.5A CN202111229346A CN113949812A CN 113949812 A CN113949812 A CN 113949812A CN 202111229346 A CN202111229346 A CN 202111229346A CN 113949812 A CN113949812 A CN 113949812A
- Authority
- CN
- China
- Prior art keywords
- state
- matrix
- motion
- image
- time
- 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
- 230000033001 locomotion Effects 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000006641 stabilisation Effects 0.000 title claims abstract description 15
- 238000011105 stabilization Methods 0.000 title claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims abstract description 51
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 19
- 230000009466 transformation Effects 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims abstract description 8
- 238000005457 optimization Methods 0.000 claims abstract description 8
- 230000004927 fusion Effects 0.000 claims abstract description 6
- 239000013598 vector Substances 0.000 claims description 10
- 238000005259 measurement Methods 0.000 claims description 7
- 238000001514 detection method Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000007704 transition Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- 238000011524 similarity measure Methods 0.000 claims description 2
- 230000001788 irregular Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 230000000903 blocking effect Effects 0.000 description 3
- 230000000087 stabilizing effect Effects 0.000 description 2
- GIYXAJPCNFJEHY-UHFFFAOYSA-N N-methyl-3-phenyl-3-[4-(trifluoromethyl)phenoxy]-1-propanamine hydrochloride (1:1) Chemical compound Cl.C=1C=CC=CC=1C(CCNC)OC1=CC=C(C(F)(F)F)C=C1 GIYXAJPCNFJEHY-UHFFFAOYSA-N 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000005311 autocorrelation function Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002924 energy minimization method Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
- H04N23/681—Motion detection
- H04N23/6811—Motion detection based on the image signal
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
- H04N23/68—Control of cameras or camera modules for stable pick-up of the scene, e.g. compensating for camera body vibrations
- H04N23/682—Vibration or motion blur correction
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
An electronic image stabilization method based on block Kalman motion prediction comprises the following steps: each image of the video stream is divided into blocks, an affine transformation matrix of each block of the adjacent inter-frame images is calculated, local motion estimation is carried out through Kalman filtering, irregular jitter in the video stream is removed, scanning motion of a camera is reserved, local motion compensation is carried out, and finally global optimization is carried out on the compensated images to complete electronic image stabilization. The invention can effectively reduce the image distortion caused by motion compensation by carrying out block description on the image, can well predict subjective motion and remove jitter by the Kalman filtering technology, and reduces the image blank area generated after motion compensation by global optimization and image fusion.
Description
Technical Field
The invention relates to an electronic image stabilizing method based on partitioned kalman motion prediction, and belongs to the technical field of image processing.
Background
Electronic Image Stabilizing Technology (EIST) is a Technology in which an observed Image sequence is unstable due to irregular motions such as vibration and shake which inevitably occur when a camera carrier shoots, and a video which is finally output to a display is a shake-blurred picture. For example, in the civil field, a navigation system, a surveillance system, a camera system, and in military applications, a surveillance and reconnaissance system, a navigation system, a guidance system, etc., video instability due to uncertain jitter of a camera carrier exists. In addition, in some use scenes, due to the fact that the working environment of the system is not ideal, the camera carrier shakes greatly, the problems that the acquired video shakes irregularly and the imaging quality is reduced are caused, the adverse effect is caused on visual application needing to acquire information or judge such as subjective observation and computer processing, the difficulty of image processing is increased, and great difficulty is caused on subsequent processing of application such as military reconnaissance, identification and monitoring.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method is characterized by overcoming the defects of the prior art, providing an electronic image stabilization method based on blocking Kalman motion prediction, particularly carrying out blocking on each image of a video stream, calculating an affine transformation matrix of each block of adjacent inter-frame images, carrying out local motion estimation through Kalman filtering, removing irregular jitter in the video stream, reserving the scanning motion of a camera, carrying out local motion compensation, and finally carrying out global optimization on the compensated image to finish electronic image stabilization.
The purpose of the invention is realized by the following technical scheme:
an electronic image stabilization method based on block Kalman motion prediction comprises the following steps:
dividing each frame of input image into a plurality of sub-blocks;
extracting and matching inter-block feature points of the reference frame and the target frame;
calculating an inter-block affine matrix;
performing motion estimation by using Kalman filtering, and then performing motion compensation;
and carrying out global optimization and image fusion so as to reduce image blank areas generated by motion compensation.
Preferably, each frame image is divided into a plurality of sub-blocks of size M x N in proportion.
Preferably, Harris angular points of each sub-block are respectively detected for a reference frame and a target frame, then two angular point sets are matched to obtain coordinate correspondence, mismatching points are removed by using a Randac algorithm to improve matching precision, coordinates are substituted into an affine motion transformation model, and an affine matrix is solved to obtain a local motion vector.
Preferably, a kalman state equation and an observation equation are set, the motion state to be predicted is introduced, a motion estimation value is obtained after prediction and update, and the motion estimation value is compared with the initial motion state to obtain a motion compensation quantity to compensate the sub-block.
Preferably, the Harris corner detection algorithm uses a differential operator to move, and introduces a Gaussian smoothing factor through Taylor series expansion, so that the noise resistance is enhanced; matching angular points in the inter-frame images by adopting similarity measurement; and removing the mismatching point pairs by adopting a random sampling consistency algorithm.
Preferably, the similarity measure is used to match the corners in the inter-frame image, and the matching criterion is:
wherein,
for the matching of the values of the corners p, q, I in the two images1,I2For the two images to be matched, (p, q) are the corner point coordinates in the two images, image I1Has a corner coordinate of (p)x,py) Image I2Has a corner coordinate of (q)x,qy)，μ1、σ1And mu2、σ2The mean and variance of pixels in a square area with the radius of D around the corner points of the two images are corresponded, and (x, y) is the coordinate point of the image pixel.
Preferably, the set kalman motion state equation and the measurement equation are as follows:
xk＝Axk-1+Buk+fk
Zk＝Hxk+vk
wherein x iskIs the system state matrix at time k, ZkIs the observed quantity of the state matrix at the time k, A is the state transition matrix, B is the control input matrix, xk-1Is k-1 orScaled system state matrix, ukTo control the input matrix at time k, fkProcess noise at time k, which is coincident with a mean of zero, and a covariance matrix of QkH is a state observation matrix, vkThe measured noise at time k is zero mean and the covariance matrix is Rk；
The state prediction equation is:
wherein,
is a state prediction value at the moment k, also called a priori state estimation value,
the covariance is estimated for the state at time k,
is a predicted value of the state at time k-1, Pk-1The covariance is estimated for the state at time k-1.
The state update equation is:
wherein, KKAt time kOptimal kalman gain, HTIn order to be a transpose of the state observation matrix H,
for state estimation at time k, PkAnd I is an identity matrix with the same dimension of the state matrix, and is an error covariance estimated value updated at the moment k.
Compared with the prior art, the invention has the following beneficial effects:
the method has the advantages that the image distortion caused by motion compensation can be effectively reduced by carrying out block description on the image, subjective motion can be well predicted and jitter can be removed through the Kalman filtering technology, and image blank areas generated after motion compensation are reduced through global optimization and image fusion.
Drawings
FIG. 1 is a flow chart of the method steps of the present invention.
Fig. 2 is a schematic diagram of image blocks of each frame.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
An electronic image stabilization method based on block Kalman motion prediction comprises the following technical solutions: although the camera carrier has irregular motion, so that irregular jitter appears in a video sequence, the camera carrier has subjective motion such as directional scanning motion, static observation motion and the like, intentional motion of a camera is estimated by calculating motion vectors between frames, unintentional motion, namely the irregular jitter, is removed, the subjective motion is kept, and therefore the video is stabilized. The main process is as follows: 1) image blocking; 2) extracting inter-block feature points and matching; 3) calculating an inter-block affine matrix; 4) local motion estimation is carried out by Kalman filtering; 5) performing motion compensation; 6) the electronic image stabilization is finally completed by global optimization and image fusion, as shown in fig. 1.
The method comprises the following steps:
1. each frame of the input image is divided into several sub-blocks of size M x N in proportion, as shown in fig. 2.
2. Detecting Harris angular points of each sub-block in a reference frame and a target frame respectively, matching the two angular point sets to obtain coordinate correspondence, removing mismatching points by using a Randac algorithm to improve matching precision, substituting the coordinates into an affine motion transformation model, and solving an affine matrix to obtain a local motion vector;
3. designing a kalman state equation and an observation equation, introducing a motion state to be predicted, predicting and updating to obtain a motion estimation value, and comparing the motion estimation value with an initial motion state to obtain a motion compensation quantity to compensate the subblock.
4. And carrying out global optimization and image fusion so as to reduce image blank areas generated by motion compensation.
Wherein, the specific process of the step two is as follows:
the Harris corner detection algorithm uses a differential operator to move, the condition of pixel gray level change in a window after the window moves any distance along any direction can be calculated through Taylor series expansion, and moreover, a Gaussian smoothing factor is introduced, so that the noise resistance is enhanced. Its autocorrelation function is:
where c (u, v) is the gray scale difference of the moving window, (u, v) is the window coordinate, (x, y) is the image pixel coordinate point, w (x, y) is the gaussian weighting function, and I (x, y) is the gray scale value at the image (x, y).
After Taylor series expansion simplification:
where M is the Harris autocorrelation matrix, IxIs a gradient value in the horizontal direction, IyIs a gradient value in the vertical direction;
according to a calculation formula of a quadratic term function characteristic value, we can solve the characteristic value of an M matrix, calculate a corner response value R to judge a corner:
R＝detM-α(traceM)2
wherein detM ═ λ1λ2To prove the determinant value of the matrix M, traceM ═ λ1+λ2Is the trace value of the matrix M, λ1And λ2The matrix M is characterized by a constant value of α, which is usually 0.04-0.06.
Then matching the corner points in the inter-image by using similarity measurement, wherein the matching criterion is as follows:
wherein,
for the corner points p, q in the two images, the value, I is matched1,I2For the two images to be matched, (p, q) are the corner point coordinates in the two images, image I1Has a corner coordinate of (p)x,py) Image I2Has a corner coordinate of (q)x,qy)，μ1、σ1And mu2、σ2The mean and variance of pixels in a square area with the radius of D around the corner points corresponding to the two images are shown.
The pairs of mismatching points are then removed using a Random Sample Consensus (RANSAC) algorithm. Classical RANSAC is a process of continuously trying different target space parameters to maximize the objective function. In the process, a subspace of the data set is randomly selected, then a model estimation can be generated, the model estimated is tested by using the rest points of the data set, a score is obtained, and finally the model estimation with the highest score is the model of the whole data set. And a threshold value is needed to be set when judging whether a certain current point is in the class.
And then, calculating an affine transformation model through the matching point pairs to obtain a local motion vector, wherein the affine transformation of the image is as follows:
wherein,
is the coordinates of pixel points before affine transformation, (x, y) is the coordinates of pixel points of images after affine transformation,
is an affine transformation matrix.
The motion compensated affine transformation model is as follows:
wherein (x)t+1，yt+1) Is the characteristic pixel point coordinate of the current frame, (x)t，yt) Is the characteristic pixel point coordinates of the reference frame,
an affine transformation matrix between the current frame and the reference frame feature points.
N affine transformation models can be obtained through n feature points obtained through registration, 6 parameters are solved by using an energy minimization method, and the following steps are carried out:
in the formula, x1，x2，……，xnAnd y1，y2，……，ynRespectively the pixel coordinates, X, of N feature points of the current frame1，X2，……，XnAnd Y1，Y2，……，YnPixel coordinates, T, of N feature points of the reference framex、TyIs the sum of the variances in the horizontal and vertical directions.
To make Tx、TyMinimum, firstly, the partial derivative is calculated, then the derivative is made to be zero, and finally, 6 parameters a to f are calculated as follows:
the specific process of the third step is as follows:
designing a kalman state equation and an observation equation, introducing a motion state to be predicted, predicting and updating to obtain a motion estimation value, and comparing the motion estimation value with an initial motion state to obtain a motion compensation quantity to compensate the subblock.
Subjective motion is extracted from the motion vector by utilizing Kalman filtering, and the motion vector of the current moment is predicted according to the motion estimation value and the covariance estimation value of the previous moment.
Establishing a kalman motion state equation and a measurement equation as follows:
xk＝Axk-1+Buk+fk
Zk＝Hxk+vk
wherein x iskIs the system state matrix at time k, ZkIs observed quantity (actual measurement) of state matrix at k time, A is state transition matrix, B is control input matrix, xk-1Is the system state matrix at time k-1, ukTo control the input matrix at time k, fkProcess noise at time k, which is coincident with a mean of zero, and a covariance matrix of QkH is a state observation matrix, vkThe measured noise at time k is zero mean and the covariance matrix is Rk。
The state prediction equation is as follows:
wherein,
is a state prediction value at the moment k, also called a priori state estimation value,
the covariance is estimated for the state at time k,
is a predicted value of the state at time k-1, Pk-1The covariance is estimated for the state at time k-1.
The state update equation is as follows:
wherein, KKFor optimal kalman gain at time k, HTIn order to be a transpose of the state observation matrix H,
for state estimation at time k, PkAnd I is an identity matrix with the same dimension of the state matrix, and is an error covariance estimated value updated at the moment k.
Obtaining motion estimation values
Then, the jitter motion vector can be obtained by comparing the jitter motion vector with the actual motion state value, and local motion compensation can be completed by compensating the jitter motion vector.
Those skilled in the art will appreciate that those matters not described in detail in the present specification are well known in the art.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to limit the present invention, and those skilled in the art can make variations and modifications of the present invention without departing from the spirit and scope of the present invention by using the methods and technical contents disclosed above.
Claims (7)
1. An electronic image stabilization method based on block Kalman motion prediction is characterized by comprising the following steps:
dividing each frame of input image into a plurality of sub-blocks;
extracting and matching inter-block feature points of the reference frame and the target frame;
calculating an inter-block affine matrix;
performing motion estimation by using Kalman filtering, and then performing motion compensation;
and carrying out global optimization and image fusion so as to reduce image blank areas generated by motion compensation.
2. The method of claim 1, wherein each frame of image is divided into a plurality of sub-blocks of size M x N in proportion.
3. The electronic image stabilization method according to claim 1, wherein Harris angular points of each sub-block are respectively detected for a reference frame and a target frame, then two angular point sets are matched to obtain coordinate correspondences, a Randac algorithm is used for removing mismatching points to improve matching accuracy, the coordinates are substituted into an affine motion transformation model, and an affine matrix is solved to obtain local motion vectors.
4. The electronic image stabilization method according to claim 1, wherein a kalman state equation and an observation equation are set, a motion state to be predicted is introduced, a motion estimation value is obtained after prediction and update, and the motion estimation value is compared with an initial motion state to obtain a motion compensation amount to compensate the sub-block.
5. The electronic image stabilization method according to claim 3, wherein the Harris corner detection algorithm uses differential operators for movement, and Gaussian smoothing factors are introduced through Taylor series expansion, so that the noise resistance is enhanced; matching angular points in the inter-frame images by adopting similarity measurement; and removing the mismatching point pairs by adopting a random sampling consistency algorithm.
6. The electronic image stabilization method according to claim 5, characterized in that similarity measures are used to match corner points in the inter-image, the matching criterion being:
wherein,
for the matching of the values of the corners p, q, I in the two images1，I2For the two images to be matched, (p, q) are the corner point coordinates in the two images, image I1Has a corner coordinate of (p)x，py) Image I2Has a corner coordinate of (q)x，qy)，μ1、σ1And mu2、σ2The mean and variance of pixels in a square area with the radius of D around the corner points of the two images are corresponded, and (x, y) is the coordinate point of the image pixel.
7. The electronic image stabilization method of claim 1, wherein:
the set kalman motion state equation and the measurement equation are as follows:
xk＝Axk-1+Buk+fk
Zk＝Hxk+vk
wherein x iskIs the system state matrix at time k, ZkIs the observed quantity of the state matrix at the time k, A is the state transition matrix, B is the control input matrix, xk-1Is the system state matrix at time k-1, ukTo control the input matrix at time k, fkProcess noise at time k, which is coincident with a mean of zero, and a covariance matrix of QkH is a state observation matrix, vkThe measured noise at time k is zero mean and the covariance matrix is Rk；
The state prediction equation is:
wherein,
is a state prediction value at the moment k, also called a priori state estimation value,
the covariance is estimated for the state at time k,
is a predicted value of the state at time k-1, Pk-1The covariance is estimated for the state at time k-1.
The state update equation is:
wherein, KKFor optimal kalman gain at time k, HTIn order to be a transpose of the state observation matrix H,
for state estimation at time k, PkAnd I is an identity matrix with the same dimension of the state matrix, and is an error covariance estimated value updated at the moment k.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111229346.5A CN113949812A (en) | 2021-10-21 | 2021-10-21 | Electronic image stabilization method based on partitioned Kalman motion prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111229346.5A CN113949812A (en) | 2021-10-21 | 2021-10-21 | Electronic image stabilization method based on partitioned Kalman motion prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113949812A true CN113949812A (en) | 2022-01-18 |
Family
ID=79331950
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111229346.5A Pending CN113949812A (en) | 2021-10-21 | 2021-10-21 | Electronic image stabilization method based on partitioned Kalman motion prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113949812A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103402045A (en) * | 2013-08-20 | 2013-11-20 | 长沙超创电子科技有限公司 | Image de-spin and stabilization method based on subarea matching and affine model |
| CN104853064A (en) * | 2015-04-10 | 2015-08-19 | 海视英科光电（苏州）有限公司 | Electronic image-stabilizing method based on infrared thermal imager |
| CN107222662A (en) * | 2017-07-12 | 2017-09-29 | 中国科学院上海技术物理研究所 | A kind of electronic image stabilization method based on improved KLT and Kalman filtering |
| CN110677578A (en) * | 2019-08-14 | 2020-01-10 | 北京理工大学 | Mixed image stabilization method and device based on bionic eye platform |
| CN110796010A (en) * | 2019-09-29 | 2020-02-14 | 湖北工业大学 | Video image stabilization method combining optical flow method and Kalman filtering |
| CN113256679A (en) * | 2021-05-13 | 2021-08-13 | 湖北工业大学 | Electronic image stabilization algorithm based on vehicle-mounted rearview mirror system |
-
2021
- 2021-10-21 CN CN202111229346.5A patent/CN113949812A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103402045A (en) * | 2013-08-20 | 2013-11-20 | 长沙超创电子科技有限公司 | Image de-spin and stabilization method based on subarea matching and affine model |
| CN104853064A (en) * | 2015-04-10 | 2015-08-19 | 海视英科光电（苏州）有限公司 | Electronic image-stabilizing method based on infrared thermal imager |
| CN107222662A (en) * | 2017-07-12 | 2017-09-29 | 中国科学院上海技术物理研究所 | A kind of electronic image stabilization method based on improved KLT and Kalman filtering |
| CN110677578A (en) * | 2019-08-14 | 2020-01-10 | 北京理工大学 | Mixed image stabilization method and device based on bionic eye platform |
| CN110796010A (en) * | 2019-09-29 | 2020-02-14 | 湖北工业大学 | Video image stabilization method combining optical flow method and Kalman filtering |
| CN113256679A (en) * | 2021-05-13 | 2021-08-13 | 湖北工业大学 | Electronic image stabilization algorithm based on vehicle-mounted rearview mirror system |
Non-Patent Citations (2)
| Title |
|---|
| CHUNG-MING KUO ET AL.: "An Efficient Motion Estimation Algorithm for Video Coding Using Kalman Filter" * |
| HAO LI ET AL.: "An Efficient Image Matching Algorithm Based on Adaptive Threshold and RANSAC" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gunjal et al. | Moving object tracking using kalman filter | |
| CN110796010B (en) | Video image stabilizing method combining optical flow method and Kalman filtering | |
| KR100985805B1 (en) | Apparatus and method for image stabilization using adaptive Kalman filter | |
| US8290212B2 (en) | Super-resolving moving vehicles in an unregistered set of video frames | |
| CN107452015B (en) | Target tracking system with re-detection mechanism | |
| CN108198201A (en) | A kind of multi-object tracking method, terminal device and storage medium | |
| JP2016508652A (en) | Determining object occlusion in image sequences | |
| US20110074927A1 (en) | Method for determining ego-motion of moving platform and detection system | |
| CN111383252B (en) | Multi-camera target tracking method, system, device and storage medium | |
| CN105374049B (en) | Multi-corner point tracking method and device based on sparse optical flow method | |
| CN108629792A (en) | Laser eyepiece detection method and device based on background modeling Yu background difference | |
| CN112580683B (en) | Multi-sensor data time alignment system and method based on cross correlation | |
| Kim et al. | Spatio-temporal weighting in local patches for direct estimation of camera motion in video stabilization | |
| CN110738667A (en) | RGB-D SLAM method and system based on dynamic scene | |
| CN116433728A (en) | DeepSORT target tracking method for shake blur scene | |
| CN117011381A (en) | Real-time surgical instrument pose estimation method and system based on deep learning and stereoscopic vision | |
| CN111160362B (en) | FAST feature homogenizing extraction and interframe feature mismatching removal method | |
| Heimbach et al. | Improving object tracking accuracy in video sequences subject to noise and occlusion impediments by combining feature tracking with Kalman filtering | |
| CN116883897A (en) | Low-resolution target identification method | |
| CN113949812A (en) | Electronic image stabilization method based on partitioned Kalman motion prediction | |
| CN115170621A (en) | Target tracking method and system under dynamic background based on relevant filtering framework | |
| Gokul et al. | Lucas Kanade based Optical Flow for Vehicle Motion Tracking and Velocity Estimation | |
| CN106934818B (en) | Hand motion tracking method and system | |
| CN114882070A (en) | Binocular vision-based three-dimensional target motion tracking method | |
| KR102173244B1 (en) | Video stabilization system based on SURF |
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 |
Application publication date: 20220118 |