CN110163894A - Sub-pixel method for tracking target based on characteristic matching - Google Patents
Sub-pixel method for tracking target based on characteristic matching Download PDFInfo
- Publication number
- CN110163894A CN110163894A CN201910397719.6A CN201910397719A CN110163894A CN 110163894 A CN110163894 A CN 110163894A CN 201910397719 A CN201910397719 A CN 201910397719A CN 110163894 A CN110163894 A CN 110163894A
- Authority
- CN
- China
- Prior art keywords
- frame image
- point
- image
- trace point
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20164—Salient point detection; Corner detection
Abstract
A kind of sub-pixel method for tracking target based on characteristic matching, comprising: the trace point of first frame image is chosen in the image continuously transmitted as benchmark trace point;First frame image and nth frame image are handled respectively, obtain the feature vector of first frame image and the feature vector of nth frame image, N is the natural number greater than 1;First frame image is matched with the feature vector of nth frame image, obtains characteristic point pair;To characteristic point to being estimated to obtain transition matrix, transition matrix and benchmark trace point are subjected to point multiplication operation, new trace point is obtained, completes the update of trace point.Sub-pixel method for tracking target proposed by the present invention based on characteristic matching can a little carry out high precision tracking in target, and have robustness when obvious characterization variation occurs for image trace point region;This method calculates simple simultaneously, can degree of parallelism it is high, be conducive to do acceleration and calculate, can be widely applied to the electro-optical countermeasure svstem of high-speed, high precision real-time tracking.
Description
Technical field
The present invention relates to image procossing, target following technical field more particularly to a kind of high-precision targets based on feature
Tracking.
Background technique
Target following is always the popular research direction of academic research and practical application in the past few decades.It is basic at present
It is divided into based on gray scale and based on the tracking of feature.Template matching and cluster two are broadly divided into track algorithm based on gray scale
Kind, these two types of method for tracking target calculate simply, also it are suitble to need the occasion of real-time tracking, but often matching error is very big,
And poor robustness.In the goal approach based on feature, wherein there is also easy drifts based on the method for tracking target of on-line study
The problems such as shifting is easily degenerated, and real-time is bad.Method for tracking target based on deep learning is hot spot at this stage, tracking accuracy
Height, but need to do large-scale training for application scenarios, the requirement of real-time is usually still not achieved.It is traditional based on feature
Method for tracking target in, target is described using feature based on the tracking of characteristic matching, passes through description vectors
Matching tracks target.Since calculating process is there are a large amount of concurrencys, this method can achieve relatively high real-time
Property.But such methods are usually all the tracking to target entirety, if often sent out for a little being tracked in target
Raw drift, or even can be with losing.
Summary of the invention
(1) technical problems to be solved
In view of this, the purpose of the present invention is to propose to a kind of sub-pixel method for tracking target based on characteristic matching, with
Phase at least is partially solved the shortcoming in above-mentioned prior art.
(2) technical solution
The present invention provides a kind of sub-pixel method for tracking target based on characteristic matching, comprising:
The trace point of first frame image is chosen in the image continuously transmitted as benchmark trace point;
First frame image and nth frame image are handled respectively, obtain the feature vector and nth frame figure of first frame image
The feature vector of picture, wherein N is the natural number greater than 1;
The feature vector of first frame image is matched with the feature vector of nth frame image, obtains characteristic point pair;
To characteristic point to being estimated to obtain transition matrix, transition matrix and benchmark trace point are subjected to point multiplication operation, obtained
To new trace point, the update of trace point is completed.
Wherein, described that the step of trace point of first frame image is as benchmark trace point is chosen in the image continuously transmitted
Include:
Determine the starting point range of objective contour in the x-direction and the z-direction;
In X, Y both direction originate the biggish direction of point range on, choose profile on two o'clock midpoint alternately with
Track point;
It is originated on the lesser direction of point range in X, Y both direction, chooses an alternative trace point every 2n pixel,
8≤2n≤10, n are natural number;
Judge in alternative trace point radius whether to have angle point in the neighborhood of n pixel, if without angle point, this is standby
Select trace point as benchmark trace point;If there is multiple alternative trace points, then randomly chooses one and be used as benchmark trace point.
Wherein, it is described obtain first frame image feature vector and nth frame image feature vector the step of include:
Contours extract is carried out to first frame image and nth frame image respectively, obtains the image block and nth frame of first frame image
The image block of image, and determine the size of the image block of first frame image and the image block of nth frame image;
The image block of the image block to first frame image and nth frame image carries out sub-pix Corner Detection respectively, obtains the
The angular coordinate of one frame image and the angular coordinate of nth frame image;
According to the angular coordinate of the angular coordinate of first frame image and nth frame image, to the angle point and N of first frame image
The angle point of frame image carries out the extraction of binary feature vector, obtains the feature vector and nth frame image of the angle point of first frame image
The feature vector of angle point.
Wherein, the contours extract step includes: image threshold, burn into expansion and contours extract;The sub-pix angle
Point detection is to be detected by Shi-Tomasi corner detection approach to described image block;The step that the binary feature vector extracts
It suddenly is to use FREAK feature extraction, and set feature vector as 256 or 516 bits.
Wherein, in the step of feature vector by first frame image is matched with the feature vector of nth frame image,
It is using Hamming distance as judgment basis, proportion threshold value is set as 8-12.
Wherein, it is described to characteristic point to being estimated in the step of obtaining transition matrix, using RANSAC algorithm to feature
Point obtains transition matrix to estimating.
Wherein, this method further includes being updated to benchmark trace point after the update for completing trace point, comprising:
One threshold value P is set, as N=P, judges that the trace point of P-1 frame image, P frame image and P+1 frame image is
It is no consistent;
If the trace point of P-1 frame image, P frame image and P+1 frame image is consistent, P frame image is set
Trace point is new benchmark trace point;
If the trace point of P-1 frame image, P frame image and P+1 frame image is inconsistent, P=P+1 is enabled, is returned
Judgment step judges whether the trace point of new P-1 frame image, P frame image and P+1 frame image is consistent, is recycled
Judgement.
(3) beneficial effect
From above-mentioned technical proposal as can be seen that a kind of sub-pixel target based on characteristic matching provided by the invention with
Track method has the advantages that
(1) the sub-pixel method for tracking target provided by the invention based on characteristic matching, by characteristic point to progress
Estimate that the operation of obtained transition matrix and benchmark trace point obtains new trace point so that target occur near trace point it is bright
When aobvious characterization changes, it is still able to achieve tenacious tracking, does not influence the robustness of algorithm.
(2) the sub-pixel method for tracking target provided by the invention based on characteristic matching, passes through sub-pixel characteristic point
Detection method and trace point matching, it is high to improve tracking accuracy.
(3) the sub-pixel method for tracking target provided by the invention based on characteristic matching, using sub-pix Corner Detection
Using hardware realization, and it can be realized faster tracking so that operand is reduced with the extraction of binary feature vector.
Detailed description of the invention
Fig. 1 is the schematic diagram of the sub-pixel method for tracking target proposed by the present invention based on characteristic matching.
Fig. 2 is the schematic diagram according to the sub-pixel method for tracking target based on characteristic matching of the embodiment of the present invention.
Fig. 3 is the schematic diagram according to the candidate trace point choosing method of the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference
Attached drawing, the present invention is described in more detail.
Fig. 1 is that the sub-pixel method for tracking target provided by the invention based on characteristic matching according to Fig. 1 should be with
Track method includes:
S101: the trace point of first frame image is chosen in the image continuously transmitted as benchmark trace point;
S102: respectively being handled first frame image and nth frame image, obtains the feature vector and the of first frame image
The feature vector of N frame image, wherein N is the natural number greater than 1;
S103: the feature vector of first frame image is matched with the feature vector of nth frame image, obtains characteristic point
It is right;
S104: to characteristic point to being estimated to obtain transition matrix, transition matrix and benchmark trace point are subjected to dot product fortune
It calculates, obtains new trace point, complete the update of trace point.
Wherein, the step of trace point of first frame image is as benchmark trace point packet is chosen in the image continuously transmitted
It includes: determining the starting point range of objective contour in the x-direction and the z-direction;The biggish side of point range is originated in X, Y both direction
Upwards, the midpoint alternately trace point of two o'clock on profile is chosen;The lesser direction of point range is originated in X, Y both direction
On, an alternative trace point is chosen every 2n pixel, 8≤2n≤10, n are natural number;Judge in alternative trace point radius as n
Whether there is angle point in the neighborhood of a pixel, if without angle point, using the alternative trace point as benchmark trace point;If there is more
A alternative trace point then randomly chooses one and is used as benchmark trace point.
Wherein, obtain first frame image feature vector and nth frame image feature vector the step of include: to first frame
Image and nth frame image carry out contours extract respectively, obtain the image block of first frame image and the image block of nth frame image, and
Determine the size of the image block of first frame image and the image block of nth frame image;Respectively to the image block and N of first frame image
The image block of frame image carries out sub-pix Corner Detection, and the angle point of the angular coordinate and nth frame image that obtain first frame image is sat
Mark;Angle point and nth frame figure according to the angular coordinate of the angular coordinate of first frame image and nth frame image, to first frame image
The angle point of picture carries out the extraction of binary feature vector, obtains the feature vector of the angle point of first frame image and the angle point of nth frame image
Feature vector.Contours extract step includes: image threshold, burn into expansion and contours extract;Sub-pix Corner Detection is logical
Shi-Tomasi corner detection approach is crossed to detect image block;The step of binary feature vector extracts is using FREAK feature
Extraction method, and feature vector is set as 256 or 516 bits.Due to herein use sub-pixel characteristic point detection method and with
Track point matching, so that tracking accuracy is high;And in conjunction with quick feature point detecting method and binary feature description vectors, make algorithm
It is more suitable for hardware realization.
Wherein, in the step of feature vector of first frame image being matched with the feature vector of nth frame image, be with
Hamming distance is judgment basis, and proportion threshold value is set as 8-12;To characteristic point to being estimated in the step of obtaining transition matrix,
Transition matrix is obtained to estimating to characteristic point using RANSAC algorithm.
Wherein, this method further includes being updated to benchmark trace point after the update for completing trace point, comprising:
One threshold value P is set, as N=P, judges that the trace point of P-1 frame image, P frame image and P+1 frame image is
It is no consistent;
If the trace point of P-1 frame image, P frame image and P+1 frame image is consistent, P frame image is set
Trace point is new benchmark trace point;
If the trace point of P-1 frame image, P frame image and P+1 frame image is inconsistent, P=P+1 is enabled, is returned
Judgment step judges whether the trace point of new P-1 frame image, P frame image and P+1 frame image is consistent, is recycled
Judgement.
Due to when target occurs significantly to characterize variation near trace point, can be realized stabilization when updating trace point
Tracking, does not influence the robustness of algorithm.
For the content that the present invention will be described in detail, spy is described with reference to the drawings for embodiment.Fig. 2 is according to of the invention real
Apply the schematic diagram of the sub-pixel method for tracking target based on characteristic matching of example.Fig. 3 is the candidate according to the embodiment of the present invention
The schematic diagram of trace point choosing method.In conjunction with Fig. 2 and Fig. 3, the tracking of the present embodiment includes 9 steps:
(1) to first frame image, i.e., present frame input picture 1 extracts objective contour, and determination will carry out feature extraction
The size of image block 3;
(2) sub-pixel Shi-Tomasi Corner Detection is carried out to image block 3;
(3) trace point is chosen from alternative trace point 15 automatically according to the characteristic distributions of angle point;
(4) angle steel joint extracts FREAK binary feature vector;
(5) to the 2nd frame image, i.e., next frame input picture 7 extracts objective contour, and determination will carry out the figure of feature extraction
As the size of block 9;
(6) sub-pixel Shi-Tomasi Corner Detection is carried out to image block 9;
(7) angle steel joint extracts FREAK binary feature vector;
(8) present frame is matched with the feature vector of angle point in next frame image block;
(9) the obtained characteristic point of matching is to being used to estimate to obtain transition matrix, and updates trace point.
As shown in Figures 2 and 3, in step 1 and step 5 contour extraction of objects using image threshold, burn into expansion,
Contours extract and etc..Specific: thresholding sets a threshold value such as 20, and image is become bianry image, and selecting radius is 2
Square structure element, to bianry image carry out etching operation, i.e., removal noise;Bianry image after etching operation is carried out
Edge extracting, edge detection operator can select Canny, the operators such as Sobel.
In addition, the tile size that each frame will carry out characteristic point detection is determined by the objective contour extracted: objective contour
Origin coordinates range in the direction x and y adds 10 (can be a range, such as 10-20), the as size of image block.
Step 2 and step 6 calculate in Shi-Tomasi Corner Detection, and the filtering of correlation matrix is having a size of 5 × 5 pixels, angle
The proportion threshold value of point is set as 0.01, and sub-pixel characteristic point coordinate is fitted to obtain using quadratic polynomial.
When step 3 chooses alternative trace point, the starting point range of objective contour in the x-direction and the z-direction is first determined;X,
It is originated on the biggish direction of point range in Y both direction, chooses the midpoint alternately trace point of two o'clock on profile;In X, Y two
It is originated on the lesser direction of point range in a direction, chooses an alternative trace point every 10 pixels;Judge alternatively tracking
Whether point radius is has angle point in the neighborhood of 5 pixels, if tracked without angle point using the alternative trace point as benchmark
Point;If there is multiple alternative trace points, then randomly chooses one and be used as benchmark trace point.Step 4 and step 7 are in FREAK feature
When extraction, feature vector is 256 or 512 bits, wherein when feature vector is 256 bit, may be implemented to track faster.
Step 8 is judgment basis with Hamming distance in characteristic matching, and proportion threshold value is set as 8.Here, Hamming distance
The number of different elements (1 or 0) in direction amount, for example the Hamming distance of vector 1011 and vector 1001 is 1.Specifically, than
Compared with the Hamming between each feature vector extracted in each feature vector extracted in current frame image and next frame image
Distance: when Hamming distance is less than threshold value, a matching double points are denoted as;When Hamming distance is greater than characteristic point, then disregard.Together
When, the smallest point of Hamming distance is selected if matching obtains multiple characteristic points pair less than threshold value for the same characteristic point
It is right.
When step 9 estimates transition matrix, estimated using RANSAC algorithm.It is not to use when present frame trace point updates
The matched jamming point of one frame is calculated, but is calculated according to benchmark trace point.Meanwhile threshold value is updated in benchmark trace point
When preceding 10 matching double points that each primary matching in front and back obtains are constant, a secondary standard trace point is updated.
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
Describe in detail it is bright, it is all at this it should be understood that be not intended to restrict the invention the above is only specific embodiments of the present invention
Within the spirit and principle of invention, any modification, equivalent substitution, improvement and etc. done should be included in protection model of the invention
Within enclosing.
Claims (10)
1. a kind of sub-pixel method for tracking target based on characteristic matching, comprising:
The trace point of first frame image is chosen in the image continuously transmitted as benchmark trace point;
First frame image and nth frame image are handled respectively, obtain the feature vector and nth frame image of first frame image
Feature vector, wherein N is the natural number greater than 1;
The feature vector of first frame image is matched with the feature vector of nth frame image, obtains characteristic point pair;
To characteristic point to being estimated to obtain transition matrix, transition matrix and benchmark trace point are subjected to point multiplication operation, obtained new
Trace point, complete the update of trace point.
2. tracking according to claim 1, which is characterized in that described to choose first frame in the image continuously transmitted
The trace point of image includes: as the step of benchmark trace point
Determine the starting point range of objective contour in the x-direction and the z-direction;
It is originated on the biggish direction of point range in X, Y both direction, chooses the midpoint alternately trace point of two o'clock on profile;
It is originated on the lesser direction of point range in X, Y both direction, every 2n pixel one alternative trace point of selection, 8≤
2n≤10, n are natural number;
Judge in alternative trace point radius whether to have angle point in the neighborhood of n pixel, if without angle point, by this alternatively with
Track point is as benchmark trace point;If there is multiple alternative trace points, then randomly chooses one and be used as benchmark trace point.
3. tracking according to claim 1, which is characterized in that the feature vector for obtaining first frame image and the
The step of feature vector of N frame image includes:
Contours extract is carried out to first frame image and nth frame image respectively, obtains the image block and nth frame image of first frame image
Image block, and determine first frame image image block and nth frame image image block size;
The image block of the image block to first frame image and nth frame image carries out sub-pix Corner Detection respectively, obtains first frame
The angular coordinate of image and the angular coordinate of nth frame image;
Angle point and nth frame figure according to the angular coordinate of the angular coordinate of first frame image and nth frame image, to first frame image
The angle point of picture carries out the extraction of binary feature vector, obtains the feature vector of the angle point of first frame image and the angle point of nth frame image
Feature vector.
4. tracking according to claim 3, which is characterized in that the contours extract step include: image threshold,
Burn into expansion and contours extract.
5. tracking according to claim 3, which is characterized in that the sub-pix Corner Detection is to pass through Shi-
Tomasi corner detection approach detects described image block.
6. tracking according to claim 3, which is characterized in that the step of binary feature vector extracts is to use
FREAK feature extraction, and feature vector is set as 256 or 516 bits.
7. tracking according to claim 1, which is characterized in that the feature vector by first frame image and N
It is using Hamming distance as judgment basis, proportion threshold value is set as 8-12 in the step of feature vector of frame image is matched.
8. tracking according to claim 1, which is characterized in that it is described to characteristic point to being estimated to obtain conversion square
In the step of battle array, transition matrix is obtained to estimating to characteristic point using RANSAC algorithm.
9. tracking according to claim 1, which is characterized in that this method is also wrapped after the update for completing trace point
It includes and benchmark trace point is updated.
10. tracking according to claim 9, which is characterized in that described to be updated to benchmark trace point, comprising:
One threshold value P is set, as N=P, judge P-1 frame image, P frame image and P+1 frame image trace point whether one
It causes;
If the trace point of P-1 frame image, P frame image and P+1 frame image is consistent, the tracking of P frame image is set
Point is new benchmark trace point;
If the trace point of P-1 frame image, P frame image and P+1 frame image is inconsistent, P=P+1 is enabled, returns to judgement
Step judges whether the trace point of new P-1 frame image, P frame image and P+1 frame image is consistent, is looped to determine.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910397719.6A CN110163894B (en) | 2019-05-14 | 2019-05-14 | Sub-pixel level target tracking method based on feature matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910397719.6A CN110163894B (en) | 2019-05-14 | 2019-05-14 | Sub-pixel level target tracking method based on feature matching |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110163894A true CN110163894A (en) | 2019-08-23 |
CN110163894B CN110163894B (en) | 2021-04-06 |
Family
ID=67634424
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910397719.6A Active CN110163894B (en) | 2019-05-14 | 2019-05-14 | Sub-pixel level target tracking method based on feature matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110163894B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111010558A (en) * | 2019-12-17 | 2020-04-14 | 浙江农林大学 | Stumpage depth map generation method based on short video image |
CN111179315A (en) * | 2019-12-31 | 2020-05-19 | 湖南快乐阳光互动娱乐传媒有限公司 | Video target area tracking method and video plane advertisement implanting method |
CN112819889A (en) * | 2020-12-30 | 2021-05-18 | 浙江大华技术股份有限公司 | Method and device for determining position information, storage medium and electronic device |
CN112819889B (en) * | 2020-12-30 | 2024-05-10 | 浙江大华技术股份有限公司 | Method and device for determining position information, storage medium and electronic device |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103400388A (en) * | 2013-08-06 | 2013-11-20 | 中国科学院光电技术研究所 | Method for eliminating Brisk (binary robust invariant scale keypoint) error matching point pair by utilizing RANSAC (random sampling consensus) |
US20150304634A1 (en) * | 2011-08-04 | 2015-10-22 | John George Karvounis | Mapping and tracking system |
CN108257155A (en) * | 2018-01-17 | 2018-07-06 | 中国科学院光电技术研究所 | A kind of extension target tenacious tracking point extracting method based on part and Global-Coupling |
CN109074657A (en) * | 2018-07-18 | 2018-12-21 | 深圳前海达闼云端智能科技有限公司 | Target tracking method and device, electronic equipment and readable storage medium |
-
2019
- 2019-05-14 CN CN201910397719.6A patent/CN110163894B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150304634A1 (en) * | 2011-08-04 | 2015-10-22 | John George Karvounis | Mapping and tracking system |
CN103400388A (en) * | 2013-08-06 | 2013-11-20 | 中国科学院光电技术研究所 | Method for eliminating Brisk (binary robust invariant scale keypoint) error matching point pair by utilizing RANSAC (random sampling consensus) |
CN108257155A (en) * | 2018-01-17 | 2018-07-06 | 中国科学院光电技术研究所 | A kind of extension target tenacious tracking point extracting method based on part and Global-Coupling |
CN109074657A (en) * | 2018-07-18 | 2018-12-21 | 深圳前海达闼云端智能科技有限公司 | Target tracking method and device, electronic equipment and readable storage medium |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111010558A (en) * | 2019-12-17 | 2020-04-14 | 浙江农林大学 | Stumpage depth map generation method based on short video image |
CN111010558B (en) * | 2019-12-17 | 2021-11-09 | 浙江农林大学 | Stumpage depth map generation method based on short video image |
CN111179315A (en) * | 2019-12-31 | 2020-05-19 | 湖南快乐阳光互动娱乐传媒有限公司 | Video target area tracking method and video plane advertisement implanting method |
CN112819889A (en) * | 2020-12-30 | 2021-05-18 | 浙江大华技术股份有限公司 | Method and device for determining position information, storage medium and electronic device |
CN112819889B (en) * | 2020-12-30 | 2024-05-10 | 浙江大华技术股份有限公司 | Method and device for determining position information, storage medium and electronic device |
Also Published As
Publication number | Publication date |
---|---|
CN110163894B (en) | 2021-04-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101887586B (en) | Self-adaptive angular-point detection method based on image contour sharpness | |
CN104268857B (en) | A kind of fast sub-picture element rim detection and localization method based on machine vision | |
CN109118523A (en) | A kind of tracking image target method based on YOLO | |
CN102222346B (en) | Vehicle detecting and tracking method | |
CN104599286B (en) | A kind of characteristic tracking method and device based on light stream | |
CN101883209B (en) | Method for integrating background model and three-frame difference to detect video background | |
CN105096299B (en) | Polygon detecting method and polygon detecting device | |
CN106023171B (en) | A kind of image angular-point detection method based on turning radius | |
CN105389807B (en) | A kind of particle filter infrared track method for merging Gradient Features and adaptive template | |
CN106920245B (en) | Boundary detection method and device | |
CN104197933B (en) | High magnitude slides enhancing and the extracting method of fixed star in a kind of range of telescope | |
CN103310453A (en) | Rapid image registration method based on sub-image corner features | |
CN101281648A (en) | Method for tracking dimension self-adaption video target with low complex degree | |
CN103268496B (en) | SAR image target recognition method | |
CN102609945B (en) | Automatic registration method of visible light and thermal infrared image sequences | |
CN104200461A (en) | Mutual information image selected block and sift (scale-invariant feature transform) characteristic based remote sensing image registration method | |
CN107146239A (en) | Satellite video moving target detecting method and system | |
CN104933738A (en) | Visual saliency map generation method based on local structure detection and contrast | |
CN110163894A (en) | Sub-pixel method for tracking target based on characteristic matching | |
CN104599291B (en) | Infrared motion target detection method based on structural similarity and significance analysis | |
Qiao et al. | Improved Harris sub-pixel corner detection algorithm for chessboard image | |
CN104200492A (en) | Automatic detecting and tracking method for aerial video target based on trajectory constraint | |
CN103761768A (en) | Stereo matching method of three-dimensional reconstruction | |
CN103913166A (en) | Star extraction method based on energy distribution | |
CN112818989A (en) | Image matching method based on gradient amplitude random sampling |
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 |