CN110008966A - One kind being based on polar quick SIFT feature point extracting method - Google Patents
One kind being based on polar quick SIFT feature point extracting method Download PDFInfo
- Publication number
- CN110008966A CN110008966A CN201910275693.8A CN201910275693A CN110008966A CN 110008966 A CN110008966 A CN 110008966A CN 201910275693 A CN201910275693 A CN 201910275693A CN 110008966 A CN110008966 A CN 110008966A
- Authority
- CN
- China
- Prior art keywords
- point
- characteristic point
- sift
- polar
- feature
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses one kind to be based on polar quick SIFT feature point extracting method, including step 1, normalized images scale space；Step 2, the screening of amendment characteristic point and positioning；Step 3 corrects characteristic point principal direction configuration method；Step 4 corrects Feature Descriptor generation method；The invention, the process for generating gradient orientation histogram to SIFT algorithm are simplified, and the distribution in the auxiliary direction of the weighted sum without histogram of gradients a, characteristic point only specifies a principal direction, reduce implementation complexity；The description that characteristic point local area image is established using polar coordinate system, establishes the Feature Descriptor of 12 × 8 dimensions, and comparison with standard method reduces implementation complexity using the Feature Descriptor of 4 × 4 × 8 dimensions of rectangular coordinate system；So that SIFT algorithm should not only be easy to hardware realization after improving, but also hardware resource consumption is less, while the basic superperformance for keeping standard SIFT algorithm.Modified core is embodied in SIFT feature detection and description algorithm.
Description
Technical field
It is specially a kind of special based on polar quick SIFT the present invention relates to intelligent technical field of image information processing
Levy point extracting method.
Background technique
With the research hotspot and difficult point that source images automatic Matching is always in intelligent Image Information Processing, need to solve
Certainly there are the accurate match in the case of larger parallax, image parallactic generally includes the rotation between image for realtime graphic and benchmark image
Turn, different scale and affine transformation etc..For target and background there are biggish otherness in infrared image, (1) is due to sensing
The variation that the factors such as different time, place, illumination condition are imaged in device will make image grayscale, contrast etc. that significant change occur,
Leading to target, there are larger gaps on the gray feature of image；(2) because the geometry of shooting angle, the introducing of shooting distance difference is abnormal
Become, the difference of imaging sensor itself resolution ratio, the factors such as system noise that imaging process introduces can all cause imaging results
It influences, this exacerbates the otherness of target in the picture.Existing matching algorithm not can solve anaglyph accurate match
Real-time after problem or algorithm engineering is difficult to meet the requirement of system application.
In recent years, the appearance and fast development of local invariant characterization method provides one to situation around cognition machint
The new technological approaches of kind.SIFT(Scale Invariant Feature Transform) method image rotation, scale become
It changes, have good invariance under affine transformation and conditions of view angle variety, can be realized rotation, scaling, brightness, relatively neglect
Characteristic matching under a variety of change conditions such as angle variation, the fuzzy, JPEG compression of image, is the solution of homologous anaglyph matching problem
Certainly one of approach.It is especially widely adopted in the scene matching aided navigation of precision guided weapon and image homing guidance, to increase
Adaptability of the strong precision guided weapon to complex environment.
However SIFT method, there is limitation, computation complexity is very high, processing data volume is larger so that in real time
It is difficult to realize.For example, need to handle the data volume for being several times as much as original image when standard SIFT algorithm extracts key point, and
The generating process of Feature Descriptor is sufficiently complex, describe son dimension it is higher cause calculation amount when characteristic matching larger, realize
When hardware resource consumption it is high, it is difficult to meet the requirement of real-time under the conditions of missile-borne.Therefore it designs a kind of based on polar quick
SIFT feature point extracting method be very it is necessary to.
Summary of the invention
The purpose of the present invention is to provide one kind to be based on polar quick SIFT feature point extracting method, above-mentioned to solve
The problem of being proposed in background technique.
In order to solve the above technical problem, the present invention provides following technical solutions: one kind being based on polar quick SIFT
Feature Points Extraction, including step 1, normalized images scale space；Step 2, the screening of amendment characteristic point and positioning；Step
Three, correct characteristic point principal direction configuration method；Step 4 corrects Feature Descriptor generation method；
Wherein in the above step 1, specific step is as follows calculating one group of image data for amendment SIFT algorithm:
1) image data forms gaussian pyramid by Gaussian smoothing；
2) DOG pyramid, DOG0-5 are formed by gaussian pyramid；
Wherein in above-mentioned steps two, amendment characteristic point screening and positioning the following steps are included:
1) Local Extremum position is found in DOG pyramid:
Each data point of 3 layers of DOG will be spatially adjacent with it data point be compared；Wherein, the data on DOG1 layer
Put 9 × 2 data points corresponding with 8 consecutive points and neighbouring level (DOG0, DOG2) of its same level totally 26 data
Point is compared, to ensure to detect the settling position of extreme point on two-dimensional space；
2) each data point of 3 layers of DOG is screened, and is accurately positioned:
It is implemented as follows: ifIt is a data point in DOG layers, 3*3 neighborhood is as follows:
Five parameters are calculated according to data vertex neighborhood:
①
②
③
④
⑤
Two intermediate results and three criterions can be calculated according to above five parameters:
⑥
⑦
⑧
⑨
⑩
Candidate point is filtered out according to three criterions:
①
2. it is linear transformation that DoG, which normalizes to 0-1, i.e.,
Criterion two is
ParameterRegulate and control parameter, default value 0.03 for contrast；
③,Default value is 10, regulates and controls parameter for principal curvatures；
Three above condition all meets, then candidate point is real extreme point, needs to export extreme point position；, i.e., extreme point first-order correction is added on the location parameter of candidate point and is exported again；
Wherein in above-mentioned steps three, the calculation method and standard SIFT in modification method gradient modulus value and direction are just the same, with
Closest to based on the corresponding Gaussian image of extreme point corresponding scale parameter σ, the neighborhood of some size is selected count
To histogram of gradients:
The range of gradient direction isDegree, wherein every 10 degree of columns, 36 columns, the peak value of histogram are then represented in total
The direction of this feature point；
Correcting the realization that characteristic point direction is distributed, steps are as follows:
1. the direction histogram containing 36 columns is generated, 0 ~ 360 degree of histogram of gradients range, wherein every 10 degree of columns, by current
The Area generation of the Gaussian image fixed size of scale, generally 7 × 7；
2. seeking characteristic point direction (as unit of 10 °):
There are three information for the characteristic point tool of image behind distribution direction: position, scale and direction；
Wherein in above-mentioned steps four, characteristic area is divided into 12 sub-regions, subregion on the basis of characteristic point principal direction
Division on the basis of principal direction, every 30 ° are a sub-regions, the pixel point set in i-th of subregion are as follows:
Centered on characteristic point, it is with characteristic point principal directionPolar coordinate system is established, whereinWithRespectively polar diameter and pole
Angle,It is characterized coordinate a little,It is characterized principal direction a little；
Then, the gradient orientation histogram in each characteristic sub-areas on 8 directions of statistics, respectively 0,45,90,135,
180,225,270,315 degree, and the gradient orientation histogram of each sub-regions is combined into one 8 × 12=96 dimension in order
Feature vector；
Subsequent weighted sum normalized and standard SIFT algorithm be it is completely the same, i.e., Gauss weighting is carried out to this vector
96 dimensional feature vectors after weighting are finally made normalized, keep it insensitive to brightness change by processing；In order to remove illumination
Influence, will be greater than after normalization 0.2 value truncation, renormalization is primary, can be further improved the distinctive of characteristic point；
Finally feature vector is normalized and generates final Feature Descriptor；Finally, carrying out the matching and correction of characteristic point.
According to the above technical scheme, in the step 1, the too big hardware resource consumption of picture size is excessive, and picture size is too
It is small to will be unable to extract effective feature point description, it is therefore desirable to comprehensively consider hardware realization difficulty, consumed resource and calculation
On the basis of method flexibility, normalized images scale.
According to the above technical scheme, the step 2 2) in, it is contemplated that the real-time and hardware resource consumption that algorithm is realized,
Seminar only carries out primary in the amendment of extreme point coordinate position, and while correcting only considers that pixel is to its position on same scale
The influence set.
According to the above technical scheme, in the step 3, correct characteristic point principal direction preparation method thought and process with
Standard SIFT algorithm is consistent, and amendment is mainly a simplified the weighting of histogram of gradients and a characteristic point is only specified
One principal direction, does not configure auxiliary direction.
According to the above technical scheme, in the step 4, the core of amendment Feature Descriptor generation method is sat using pole
Mark system establishes the description of characteristic point local area image.
According to the above technical scheme, in the step 4, on the basis of characteristic point principal direction by characteristic area divide this
Kind region division is different from standard SIFT algorithm, it is advantageous that not needing the rotation of progress image, can greatly simplify feature
The complexity that description generates.
Compared with prior art, the beneficial effects obtained by the present invention are as follows being: polar quick SIFT feature should be based on
Extracting method, the process for generating gradient orientation histogram to SIFT algorithm are simplified, the weighting without histogram of gradients
With the distribution in auxiliary direction, a characteristic point only specifies a principal direction, reduces implementation complexity；It is established using polar coordinate system special
The description of sign point local area image, establishes the Feature Descriptor of 12 × 8 dimensions, comparison with standard method is using rectangular coordinate system
The Feature Descriptor of 4 × 4 × 8 dimensions reduces implementation complexity.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is infrared homologous images match schematic diagram of the invention；
Fig. 2 is picture breakdown pyramid schematic diagram of the invention；
Fig. 3 is the subcharacter region division schematic diagram under polar coordinate system of the invention；
Fig. 4 is original image in test image l-G simulation test one of the invention；
Fig. 5 is matching result schematic diagram after rotating counterclockwise in test image l-G simulation test one of the invention；
Fig. 6 is original image in test image l-G simulation test two of the invention；
Fig. 7 is that rear matching result schematic diagram is rotated clockwise in test image l-G simulation test two of the invention；
Fig. 8 is original image in test image l-G simulation test three of the invention；
Fig. 9 is that matching result schematic diagram after noise is added in test image l-G simulation test three of the invention；
Figure 10 is original image in test image l-G simulation test four of the invention；
Figure 11 is matching result schematic diagram after four mesoscale of test image l-G simulation test scaling and rotation of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1-11 is please referred to, the present invention provides a kind of technical solution: one kind is mentioned based on polar quick SIFT feature
Take method:
Test image l-G simulation test one: rotation counterclockwise
Such as Fig. 4, the image to be matched of selection is the 1st frame and the 8th frame of one group of sequence image, and the 8th frame image is along counterclockwise
Direction has rotated 37 degree, and matching result is as shown in figure 5, it can be seen that improved SIFT algorithm can obtain more
With point, there is robustness to the rotation of image.
Test image l-G simulation test two: it rotates clockwise
Such as Fig. 6, the 8th frame image 17 degree are rotated in a clockwise direction, matching result is as shown in fig. 7, it can be seen that change
SIFT algorithm after can obtain more match point.
Test image l-G simulation test three: noise
Such as Fig. 8, being added to mean value to the 8th frame image is 0, the Gaussian noise that standard deviation is 0.1, matching result as shown in figure 9, from
In it can be seen that improved SIFT algorithm can obtain more match point, the Gaussian noise of the algorithm pattern image has certain
Adaptability.
Test image l-G simulation test four: scaling and rotation
Such as Figure 10, if not only there is the rotations of larger angle between image, but also there are certain scalings, and matching result is such as
Shown in Figure 11, it can be seen that amendment SIFT algorithm still is able to correctly match.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.
Finally, it should be noted that the foregoing is only a preferred embodiment of the present invention, it is not intended to restrict the invention,
Although the present invention is described in detail referring to the foregoing embodiments, for those skilled in the art, still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features.
All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in of the invention
Within protection scope.
Claims (6)
1. one kind is based on polar quick SIFT feature point extracting method, including step 1, normalized images scale space；Step
Rapid two, amendment characteristic point is screened and positioning；Step 3 corrects characteristic point principal direction configuration method；Step 4, amendment feature description
Sub- generation method；It is characterized by:
Wherein in the above step 1, specific step is as follows calculating one group of image data for amendment SIFT algorithm:
1) image data forms gaussian pyramid by Gaussian smoothing；
2) DOG pyramid, DOG0-5 are formed by gaussian pyramid；
Wherein in above-mentioned steps two, amendment characteristic point screening and positioning the following steps are included:
1) Local Extremum position is found in DOG pyramid:
Each data point of 3 layers of DOG will be spatially adjacent with it data point be compared；Wherein, the data on DOG1 layer
Put 9 × 2 data points corresponding with 8 consecutive points and neighbouring level (DOG0, DOG2) of its same level totally 26 data
Point is compared, to ensure to detect the settling position of extreme point on two-dimensional space；
2) each data point of 3 layers of DOG is screened, and is accurately positioned:
It is implemented as follows: ifIt is a data point in DOG layers, 3*3 neighborhood is as follows:
Five parameters are calculated according to data vertex neighborhood:
①
②
③
④
⑤
Two intermediate results and three criterions can be calculated according to above five parameters:
⑥
⑦
⑧
⑨
⑩
Candidate point is filtered out according to three criterions:
①
2. it is linear transformation that DoG, which normalizes to 0-1, i.e.,
Criterion two is
ParameterRegulate and control parameter, default value 0.03 for contrast；
③,Default value is 10, regulates and controls parameter for principal curvatures；
Three above condition all meets, then candidate point is real extreme point, needs to export extreme point position；,
Extreme point first-order correction is added on the location parameter of candidate point and is exported again；
Wherein in above-mentioned steps three, the calculation method and standard SIFT in modification method gradient modulus value and direction are just the same, with
Closest to based on the corresponding Gaussian image of extreme point corresponding scale parameter σ, the neighborhood of some size is selected count
To histogram of gradients:
The range of gradient direction isDegree, wherein every 10 degree of columns, 36 columns, the peak value of histogram are then represented in total
The direction of this feature point；
Correcting the realization that characteristic point direction is distributed, steps are as follows:
1. the direction histogram containing 36 columns is generated, 0 ~ 360 degree of histogram of gradients range, wherein every 10 degree of columns, by current
The Area generation of the Gaussian image fixed size of scale, generally 7 × 7；
2. seeking characteristic point direction (as unit of 10 °):
There are three information for the characteristic point tool of image behind distribution direction: position, scale and direction；
Wherein in above-mentioned steps four, characteristic area is divided into 12 sub-regions, subregion on the basis of characteristic point principal direction
Division on the basis of principal direction, every 30 ° are a sub-regions, the pixel point set in i-th of subregion are as follows:
Centered on characteristic point, it is with characteristic point principal directionPolar coordinate system is established, whereinWithRespectively polar diameter and pole
Angle,It is characterized coordinate a little,It is characterized principal direction a little；
Then, the gradient orientation histogram in each characteristic sub-areas on 8 directions of statistics, respectively 0,45,90,135,
180,225,270,315 degree, and the gradient orientation histogram of each sub-regions is combined into one 8 × 12=96 dimension in order
Feature vector；
Subsequent weighted sum normalized and standard SIFT algorithm be it is completely the same, i.e., Gauss weighting is carried out to this vector
96 dimensional feature vectors after weighting are finally made normalized, keep it insensitive to brightness change by processing；In order to remove illumination
Influence, will be greater than after normalization 0.2 value truncation, renormalization is primary, can be further improved the distinctive of characteristic point；
Finally feature vector is normalized and generates final Feature Descriptor；Finally, carrying out the matching and correction of characteristic point.
2. according to claim 1 a kind of based on polar quick SIFT feature point extracting method, it is characterised in that: institute
It states in step 1, the too big hardware resource consumption of picture size is excessive, and picture size is too small to be will be unable to extract effective characteristic point
Description, it is therefore desirable on the basis of comprehensively considering hardware realization difficulty, consumed resource and algorithm flexibility, normalized images
Scale.
3. according to claim 1 a kind of based on polar quick SIFT feature point extracting method, it is characterised in that: institute
State step 2 2) in, it is contemplated that the real-time and hardware resource consumption that algorithm is realized, seminar repairs extreme point coordinate position
Influence of the pixel to its position on same scale is only considered when just above only carrying out once, and correcting.
4. according to claim 1 a kind of based on polar quick SIFT feature point extracting method, it is characterised in that: institute
It states in step 3, the thought and process for correcting characteristic point principal direction preparation method are consistent with standard SIFT algorithm, are corrected
A principal direction is only specified in the weighting and a characteristic point for being mainly a simplified histogram of gradients, does not configure auxiliary direction.
5. according to claim 1 a kind of based on polar quick SIFT feature point extracting method, it is characterised in that: institute
It states in step 4, the core of amendment Feature Descriptor generation method establishes characteristic point local area image using polar coordinate system
Description.
6. according to claim 1 a kind of based on polar quick SIFT feature point extracting method, it is characterised in that: institute
It states in step 4, this region division and standard SIFT algorithm for being divided characteristic area on the basis of characteristic point principal direction are not
Together, it is advantageous that not needing the rotation of progress image, the complexity of Feature Descriptor generation can be greatly simplified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910275693.8A CN110008966A (en) | 2019-04-08 | 2019-04-08 | One kind being based on polar quick SIFT feature point extracting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910275693.8A CN110008966A (en) | 2019-04-08 | 2019-04-08 | One kind being based on polar quick SIFT feature point extracting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110008966A true CN110008966A (en) | 2019-07-12 |
Family
ID=67170252
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910275693.8A Pending CN110008966A (en) | 2019-04-08 | 2019-04-08 | One kind being based on polar quick SIFT feature point extracting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110008966A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275053A (en) * | 2020-01-16 | 2020-06-12 | 北京联合大学 | Method and system for representing local feature descriptor |
| WO2021017361A1 (en) * | 2019-07-31 | 2021-02-04 | 苏州中科全象智能科技有限公司 | Template matching algorithm based on edge and gradient feature |
| CN114596347A (en) * | 2022-03-14 | 2022-06-07 | 西南交通大学 | Three-dimensional reconstruction and volume calculation method for landslide mass based on mobile photographic image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120182442A1 (en) * | 2011-01-14 | 2012-07-19 | Graham Kirsch | Hardware generation of image descriptors |
| CN102902973A (en) * | 2012-09-28 | 2013-01-30 | 中国科学院自动化研究所 | Extraction method of image characteristic with rotation invariance |
| CN102945289A (en) * | 2012-11-30 | 2013-02-27 | 苏州搜客信息技术有限公司 | Image search method based on CGCI-SIFT (consistence index-scale invariant feature transform) partial feature |
| CN104156696A (en) * | 2014-07-23 | 2014-11-19 | 华南理工大学 | Bi-directional-image-based construction method for quick local changeless feature descriptor |
-
2019
- 2019-04-08 CN CN201910275693.8A patent/CN110008966A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120182442A1 (en) * | 2011-01-14 | 2012-07-19 | Graham Kirsch | Hardware generation of image descriptors |
| CN102902973A (en) * | 2012-09-28 | 2013-01-30 | 中国科学院自动化研究所 | Extraction method of image characteristic with rotation invariance |
| CN102945289A (en) * | 2012-11-30 | 2013-02-27 | 苏州搜客信息技术有限公司 | Image search method based on CGCI-SIFT (consistence index-scale invariant feature transform) partial feature |
| CN104156696A (en) * | 2014-07-23 | 2014-11-19 | 华南理工大学 | Bi-directional-image-based construction method for quick local changeless feature descriptor |
Non-Patent Citations (4)
| Title |
|---|
| 唐永鹤等: "基于Laplacian的局部特征描述算法", 《光学精密工程》 * |
| 唐永鹤等: "基于灰度差分不变量的快速局部特征描述算法", 《光学精密工程》 * |
| 林陶等: "尺度不变特征转换算法在图像特征提取中的应用", 《计算机应用》 * |
| 粼粼淇: "SIFT特征点提取", 《CSDN:HTTPS://BLOG.CSDN.NET/LINGYUNXIANHE/ARTICLE/DETAILS/79063547》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021017361A1 (en) * | 2019-07-31 | 2021-02-04 | 苏州中科全象智能科技有限公司 | Template matching algorithm based on edge and gradient feature |
| CN111275053A (en) * | 2020-01-16 | 2020-06-12 | 北京联合大学 | Method and system for representing local feature descriptor |
| CN111275053B (en) * | 2020-01-16 | 2023-11-10 | 北京腾信软创科技股份有限公司 | Method and system for representing local feature descriptors |
| CN114596347A (en) * | 2022-03-14 | 2022-06-07 | 西南交通大学 | Three-dimensional reconstruction and volume calculation method for landslide mass based on mobile photographic image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105957015B (en) | A kind of 360 degree of panorama mosaic methods of threaded barrel inner wall image and system | |
| CN107506763B (en) | Multi-scale license plate accurate positioning method based on convolutional neural network | |
| US8553980B2 (en) | Method and apparatus extracting feature points and image based localization method using extracted feature points | |
| CN109859226B (en) | Detection method of checkerboard corner sub-pixels for graph segmentation | |
| CN110008966A (en) | One kind being based on polar quick SIFT feature point extracting method | |
| CN110992263B (en) | Image stitching method and system | |
| CN105224949B (en) | SAR image terrain classification method based on cross-cutting transfer learning | |
| CN105096317A (en) | Fully automatic calibration method for high performance camera under complicated background | |
| CN109961485A (en) | A method of target positioning is carried out based on monocular vision | |
| Urban et al. | Finding a good feature detector-descriptor combination for the 2D keypoint-based registration of TLS point clouds | |
| CN104616297A (en) | Improved SIFI algorithm for image tampering forensics | |
| WO2010150639A1 (en) | Method and device for determining shape congruence in three dimensions | |
| CN104240231A (en) | Multi-source image registration based on local structure binary pattern | |
| CN108229500A (en) | A kind of SIFT Mismatching point scalping methods based on Function Fitting | |
| CN104123554A (en) | SIFT image characteristic extraction method based on MMTD | |
| CN114331879A (en) | Visible light and infrared image registration method for equalized second-order gradient histogram descriptor | |
| CN108983769B (en) | Instant positioning and map construction optimization method and device | |
| CN110443295A (en) | Improved images match and error hiding reject algorithm | |
| CN111680704A (en) | Automatic and rapid extraction method and device for newly-increased human active plaque of ocean red line | |
| CN111861866A (en) | Panoramic reconstruction method for substation equipment inspection image | |
| CN109003307A (en) | Fishing mesh sizing method based on underwater Binocular vision photogrammetry | |
| CN117576219A (en) | Camera calibration equipment and calibration method for single shot image of large wide-angle fish-eye lens | |
| CN110929782A (en) | River channel abnormity detection method based on orthophoto map comparison | |
| Afzali et al. | Foreground and background feature fusion using a convex hull based center prior for salient object detection | |
| CN112418226A (en) | Method and device for identifying opening and closing states of fisheyes |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190712 |