CN106067175B - A kind of solar activity object detection method based on square net structure - Google Patents
A kind of solar activity object detection method based on square net structure Download PDFInfo
- Publication number
- CN106067175B CN106067175B CN201610374018.7A CN201610374018A CN106067175B CN 106067175 B CN106067175 B CN 106067175B CN 201610374018 A CN201610374018 A CN 201610374018A CN 106067175 B CN106067175 B CN 106067175B
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- threshold
- area
- row
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of solar activity object detection methods based on square net structure, belong to astronomical technology and technical field of image processing.The present invention includes the following steps：Image normalization：Reduce the tonal gradation of original image；Square net structure is defined, and divides image；The rectangular coordinate system of save mesh image is established, and to grid cell dimensionality reduction and is stored；Estimate initial threshold using histogram thresholding method；Using minimum extraneous rectangle as tool correction threshold, Grads threshold is obtained；Identification object region hides background area；Divide and preserve in target area.The present invention compared with the existing technology, realizes the image segmentation of a plurality of types of solar activitys, also to solve the problems, such as that magnanimity chronometer data storage provides a kind of feasible solution.
Description
Technical field
The present invention relates to a kind of solar activity object detection methods based on square net structure, belong to astronomical technology and figure
As processing technology field.
Background technology
In astronomical research field, FITS formats is generally used to store astronomical image.With the astronomical picture quality absorbed
Higher and higher, the astronomical big order of magnitude storage of image just becomes a stubborn problem.Existing many image partition methods are still
It does not solve the problem above-mentioned, because segmentation result still retains incoherent background area, data set size does not become significantly
Change, these will increase the space-time expense of late feature extraction.
Identify that the method for solar activity mainly has seed region growth method, watershed algorithm and threshold method on single image
Etc. several.Although current recognizer combines the physical features of solar activity, and integrated use image procossing is various
Technology, therefore the identification on single width figure achieves relatively good as a result, for example, region-growing method and dividing ridge method are because it is each
It is widely used in solar activity segmentation from feature, has good recognition effect for sunspot and the filaments of sun, these sun are lived
It is dynamic that there is apparent gray feature.But corona sprays, the solar activitys such as filament activation, gray feature is with respect to unobvious, area
Domain growth method and dividing ridge method can cause two kinds of extreme knots of over-segmentation or image cavity to the segmentation of the above solar activity
Fruit.There is presently no current methods to be identified for a variety of solar activitys on full-time face image.
The present invention proposes a kind of solar activity target detection side based on square net structure in order to solve these problems
Zone of action can be saved as new data set by method, this method after dividing solar activity.First, this method constructs one
Storage unit of the kind based on square net is used to store the gray value of image, then threshold value Selection Strategy is utilized to be selected from threshold interval
Suitable threshold value is selected, realizes that different threshold segmentation methods divide different solar activitys.
Invention content
The present invention provides a kind of solar activity object detection methods based on square net structure, realize multiple types
Solar activity image segmentation, also to solve the problems, such as that magnanimity chronometer data storage provides a kind of feasible solution.
Technical scheme of the present invention includes image preprocessing and solar activity target detection two parts：First, to original
FITS images are normalized, and reduce tonal gradation；Secondly, mesh generation is carried out to image using the network of definition；Again
Person establishes the rectangular coordinate system of save mesh image, to image dimensionality reduction and storage；Furthermore estimated just using histogram thresholding method
Beginning threshold value；Furthermore using minimum enclosed rectangle as tool correction threshold, obtain Grads threshold；Furthermore identification and separation target
Region hides background area；Finally, divide and preserve target area.
A kind of solar activity object detection method based on square net structure is as follows：
Step 1：Original sun image is normalized；
Step 2：Mesh generation is carried out to image using the network of definition；
Step 3：The rectangular coordinate system for establishing save mesh image, to image dimensionality reduction and storage；
Step 4：Initial threshold is determined using histogram thresholding method；
Step 5：Using minimum enclosed rectangle as tool correction threshold, Grads threshold is obtained；
Step 6：Identification and separation target area, background area is hidden；
Step 7：Segmentation and preservation target area.
Original sun image is normalized using mean filter in the step 1, by gradation of image grade from
16 potential drops are to 8.
Network is defined as follows in the step 2：
Wherein, [M, N] indicates original image size, is M rows, N row, [m, n] indicates the size of each grid image, INT
Function judges whether number is shaping,WithRespectively indicate M and N open square root be integer as a result,
Floor (M) and floor (N) indicates the integer no more than M and N respectively, and SQRT [floor (M)] and SQRT [floor (N)] are respectively
Indicate the integer that can be opened quadratic power no more than M and N.
The rectangular coordinate system that save mesh image is established in the step 3 is as follows to the process of image dimensionality reduction and storage：
For original image using four-dimensional storage of array, structure is as follows：
Block(pixel_row,pixel_col,cell_row,cell_col)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Cell_row indicates that the line number where grid cell, cell_col indicate the columns where grid cell；
Dimension-reduction treatment stores image using three-dimensional array, and structure is as follows：
Block(pixel_row,pixel_col,index)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Index indicates coordinate of the grid cell in the one-dimensional space.
Determine that the process of initial threshold is as follows in the step 4：
Visual signature based on human eye is divided into sightless background area, it is seen that dark in region to whole image
Brighter areas in region, visibility region, corresponding gray scale interval are respectively 0~99,100~150,151~255, establish ash
Histogram is spent, histogram thresholding method statistics determines initial gray threshold value T0It is 160~185.
The detailed process that Grads threshold is arranged in the step 5 is as follows：
According to above-mentioned square net structure and initial threshold T0, each grid cell is split, grid list is utilized
The minimum enclosed rectangle correction threshold of member, even minimum enclosed rectangle area and the ratio of image area is α, if α≤area_
Threshold sets new threshold value T1(T1<T0), if α ＞ area_threshold, set new threshold value T2(T2>T0), wherein
The process that target area is identified and detached in the step 6 is as follows：
First, it counts one by one and is more than threshold value T (T=in each grid Block (pixel_row, pixel_col, index)
T1Or T2) number of pixels num retain target area, setting grid cell Block if num >=num_threshold
(pixel_row, pixel_col, index) all pixels value is 255, and wherein num_threshold takes empirical value 0.15；
Then, the density d ensity for counting the grid cell of all connections, refer to the grid cell of connection area summation with
The ratio of connected region minimum enclosed rectangle area deletes background area if density≤density_threshold, i.e., will
These all pixels values for meeting the grid cell Block (pixel_row, pixel_col, index) of condition are 0, wherein
Density_threshold values 0.5；
Finally, target area is marked out with minimum enclosed rectangle, the wherein solution of minimum enclosed rectangle is established in square grid
On the basis of lattice structure, the gap code for the boundary for finding out target area is first had to：{(u0,v0)d0d1d2…dn-1}
Wherein, Pk∈{P0,P1,P2…Pn-1The each salient point of storage coordinate value, P0=(u0,v0) indicate starting point seat
Mark, n indicate the number on the pixel side of target area outermost；d0Indicate starting point P0To second salient point P1Mobile direction, d1Table
Show salient point P1To third salient point P2Mobile direction, dn-1Indicate salient point Pn-1To starting point P0Mobile direction, d0, d1... dn-1
∈ { 0,1,2,3 }, wherein it is to the right, upwards, to the left and downwards, in turn by gap code that { 0,1,2,3 } corresponds to moving direction respectively
Find out four apex coordinates of the minimum enclosed rectangle of target area.
The beneficial effects of the invention are as follows：
1, using square net structure as minimum treat unit, make to accelerate the processing procedure of image；For different zones
Grid choose different threshold values, improve processing accuracy；The effective interference for preventing picture noise.
2, a variety of solar activitys are directed to, the target area of solar activity can be preferably partitioned into.
3, the target area of solar activity and background area are detached, and stores target area, effectively reduce data
The size of collection.
Description of the drawings
Fig. 1 is the overview flow chart of the present invention；
Fig. 2 be in the present invention solar dynamics observatory (SDO) in 1600 full resolution prictures that G-band observes
One figure；
Fig. 3 is the present invention to Fig. 2 figures after pretreatment；
Fig. 4 is the result figure that the present invention carries out Fig. 3 mesh generation；
Fig. 5 is that the present invention carries out foreground and background separation to Fig. 4；
Fig. 6 is result figure of the present invention to solar activity zone marker in Fig. 5；
Fig. 7 is that the present invention carries out Fig. 6 the result figure after image segmentation.
Specific implementation mode
Below with reference to the embodiments and with reference to the accompanying drawing being described further to technical scheme of the present invention.
Embodiment：Such as Fig. 1-7, a kind of solar activity object detection method based on square net structure, first, to original
FITS images are normalized, and reduce tonal gradation；Secondly, mesh generation is carried out to image using the network of definition；Again
Person establishes the rectangular coordinate system of save mesh image, to image dimensionality reduction and storage；Furthermore estimated just using histogram thresholding method
Beginning threshold value；Furthermore using minimum enclosed rectangle as tool correction threshold, obtain Grads threshold；Furthermore identification and separation target
Region hides background area；Finally, divide and preserve target area.
The solar activity object detection method based on square net structure is as follows：
Step 1：Image preprocessing：Firstly, since the highest series of eye recognition degree be 8 (tonal gradation 0~
255), image is normalized using mean filter, by tonal gradation from 16 potential drops to 8, original image such as Fig. 2,
The results are shown in Figure 3 after normalized.
Then, the rectangular coordinate system of grid image is established, network is defined：
[M, N] indicates that original image size is M rows, N row.[m, n] then represents the size of each grid image, provides m, n
Solution：
Wherein, INT functions judge whether number is shaping,Indicate that M (N) opens square root
As a result it is integer, floor (M) (floor (N)) indicates the integer no more than M (N), SQRT [floor (M)] (SQRT [floor
(N)] integer that can be opened quadratic power no more than M (N)) is indicated.Image is divided using network, the results are shown in Figure 4.
Then, conventional Meshing Method stores image using four-dimensional structure of arrays, and space expense is larger, builds here
The rectangular coordinate system of vertical grid image, horizontal direction is as unit of the columns of square net, and vertical direction is with the row of square net
Number is that unit refers here to the number that four-dimensional array turns three-dimensional array using three-dimensional array structure storage image (dimensionality reduction storage)
According to Structural Transformation.
For original image using four-dimensional storage of array, structure is as follows：
Block(pixel_row,pixel_col,cell_row,cell_col)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Cell_row indicates that the line number where grid cell, cell_col indicate the columns where grid cell.Dimension-reduction treatment, using three
Storage of array image is tieed up, structure is as follows：
Block(pixel_row,pixel_col,index)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Index indicates coordinate of the grid cell in the one-dimensional space.
Step 2：Set initial threshold：First, the visual signature based on human eye can define three kinds of visualization ranks：It can not
The background area (gray scale interval 0~99) seen, it is seen that the darker area (gray scale interval 100~150) in region, it is seen that in region
Brighter areas (gray scale interval 151~255) together constitute visible target area.Then, histogram thresholding method is united first
The grey level histogram for counting 1600 images, in 1600 unimodal histograms of statistical result showed, shared by threshold interval 100~150
Ratio is higher than 60%, and 151~255 proportions are less than 40%,
Peak value is between 100~150, it is considered that " shoulder " of waveform is the marginal point of foreground and background, i.e. threshold value
Selected point.It is shown, ranging from 160~185 where the waveform " shoulder " of 1600 histograms, is chosen here initial by statistics
Threshold value T0It is 175.
Step 3：Grads threshold is set：Single threshold split plot design is common methods, and the present invention uses grads threshold method.Meet
In the case of different decision condition, different threshold values is distributed.According to above-mentioned square net structure and initial threshold T0=175,
Tentatively each grid cell is split, using the minimum enclosed rectangle correction threshold of grid cell, (minimum enclosed rectangle is
Most common method in image classification and Study of recognition is used to the geometric properties of extraction target area).Made according to the present invention
The characteristics of with image set, it is [0.3,0.7] that the effective coverage area of solar activity, which accounts for artwork ratio, in this, as empirical value.
Grads threshold method process is as follows：
Decision condition 1：If minimum enclosed rectangle area and the ratio of image area are α, if α≤area_threshold,
Set new threshold value T1(T1<T0)。
Decision condition 2：If α ＞ area_threshold set new threshold value T2(T2>T0)。
HereIf α≤0.3, illustrate that divided area is too small, sets new threshold
Value T1=150；If α ＞ 0.7 illustrate that divided area is excessive, new threshold value T is set2=200.
Step 4：Target area identifies：Based on network, divides image using threshold method, be first depending in step 1 and build
Vertical rectangular coordinate system counts be more than threshold value T (T in each grid Block (pixel_row, pixel_col, index) one by >1Or T2) number of pixels num.Process is as follows：
Decision condition 3：If num >=num_threshold, retain target area, that is, grid cell Block is set
The all pixels value of (pixel_row, pixel_col, index) is 255.
Decision condition 4：The density d ensity for counting the grid cell of all connections, if density≤density_
Threshold, delete background area, i.e., by these meet condition grid cell Block (pixel_row, pixel_col,
Index all pixels value) is 0.
Here the density of grid cell refers to：Area summation and the connected region minimum enclosed rectangle of the grid cell of connection
The ratio of area, num_threshold take empirical value 0.15, density_threshold to take empirical value 0.5.
Finally, target area is marked out with minimum enclosed rectangle.Wherein the solution of minimum enclosed rectangle is established in square grid
On the basis of lattice structure, the gap code for the boundary for finding out target area is first had to：
{(u0,v0)d0d1d2…dn-1}
Wherein, Pk∈{P0,P1,P2…Pn-1The each salient point of storage coordinate value, P0=(u0,v0) indicate starting point seat
Mark, n indicate the number on the pixel side of target area (white portion in Fig. 5) outermost；d0Indicate starting point P0To second salient point P1
Mobile direction, d1Indicate salient point P1To third salient point P2Mobile direction, dn-1Indicate salient point Pn-1To starting point P0Mobile
Direction, d0, d1... dn-1∈ { 0,1,2,3 }, wherein { 0,1,2,3 } correspond to respectively moving direction be to the right, upwards, to the left and to
Under, find out four apex coordinates of the minimum enclosed rectangle of target area in turn by gap code.
Target area and background area, target area area and original image area ratio α=35% are isolated in Fig. 5,
The visibility region ratio of intensity histogram graph discovery after comparison segmentation, tonal range 151~255 increases, tonal range 0~150
Area ratio reduce, the results showed that after segmentation, the target area of each solar activity is effectively identified, background area quilt
It hides, this has benefited from being used in combination for square net structure and Grads threshold.Meanwhile square net structure not only increases image
The precision of segmentation, meanwhile, make to accelerate the processing procedure of image；The effective interference for preventing picture noise.
Step 5：Divide target area：The solar activity of target area is saved as to new datagram image set.By target area
The solar activity in domain saves as new datagram image set.By the bianry image of Fig. 6 and original FITS image superpositions, obtain including mesh
Mark the new images in region.The 50% of all images are accounted for for quantity of the ration of division in [0.3,0.7] of eight kinds of solar activitys,
The size of data set is greatly simplified.
The specific implementation mode of the present invention is explained in detail above in conjunction with attached drawing, but the present invention is not limited to above-mentioned
Embodiment within the knowledge of a person skilled in the art can also be before not departing from present inventive concept
Put that various changes can be made.
Claims (3)
1. a kind of solar activity object detection method based on square net structure, it is characterised in that：Include the following steps：
Step 1：Original sun image is normalized；
Step 2：Mesh generation is carried out to image using the network of definition；
Step 3：The rectangular coordinate system for establishing save mesh image, to image dimensionality reduction and storage, process is as follows：
For original image using four-dimensional storage of array, structure is as follows：
Block(pixel_row,pixel_col,cell_row,cell_col)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Cell_row indicates that the line number where grid cell, cell_col indicate the columns where grid cell；
Dimension-reduction treatment stores image using three-dimensional array, and structure is as follows：
Block(pixel_row,pixel_col,index)
Wherein, pixel_row indicates that the line number where gray shade unit, pixel_col indicate the columns where gray shade unit,
Index indicates coordinate of the grid cell in the one-dimensional space；
Step 4：Initial threshold is determined using histogram thresholding method；
Step 5：Using minimum enclosed rectangle as tool correction threshold, Grads threshold is obtained, detailed process is as follows：
According to above-mentioned square net structure and initial threshold T0, each grid cell is split, most using grid cell
Small boundary rectangle correction threshold, even minimum enclosed rectangle area and the ratio of image area are α, if α≤area_
Threshold sets new threshold value T1, T1<T0If α ＞ area_threshold set new threshold value T2, T2>T0, wherein
Wherein determine that the process of initial threshold is as follows：
Visual signature based on human eye, whole image is divided into darker area in sightless background area, visibility region,
Brighter areas in visibility region, corresponding gray scale interval are respectively 0~99,100~150,151~255, establish intensity histogram
Figure, histogram thresholding method statistics determine initial threshold T0It is 160~185；
Step 6：Identification and separation target area, background area are hidden, process is as follows：
First, the pixel in each grid Block (pixel_row, pixel_col, index) more than threshold value T is counted one by one
Number num, T=T1Or T=T2If num >=num_threshold, retain target area, setting grid cell Block (pixel_
Row, pixel_col, index) all pixels value be 255, wherein num_threshold takes empirical value 0.15；
Then, the density d ensity for counting the grid cell of all connections, the area for referring to the grid cell of connection are summed and are connected to
The ratio of region minimum enclosed rectangle area deletes background area, i.e., by these if density≤density_threshold
The all pixels value for meeting the grid cell Block (pixel_row, pixel_col, index) of condition is set as 0, wherein
Density_threshold values 0.5；
Finally, target area is marked out with minimum enclosed rectangle, the wherein solution of minimum enclosed rectangle is established in square net knot
On the basis of structure, the gap code for the boundary for finding out target area is first had to：
{(u0,v0)d0d1d2…dn-1}
Wherein, Pk∈{P0,P1,P2…Pn-1The each salient point of storage coordinate value, P0=(u0,v0) indicate starting point coordinate, n tables
Show the number on the pixel side of target area outermost；d0Indicate starting point P0To second salient point P1Mobile direction, d1Indicate salient point
P1To third salient point P2Mobile direction, dn-1Indicate salient point Pn-1To starting point P0Mobile direction, d0, d1... dn-1∈{0,
1,2,3 }, wherein it is to the right, upwards, to the left and downwards, by gap code and then to find out mesh that { 0,1,2,3 } corresponds to moving direction respectively
Mark four apex coordinates of the minimum enclosed rectangle in region；
Step 7：Segmentation and preservation target area.
2. the solar activity object detection method according to claim 1 based on square net structure, it is characterised in that：Institute
It states in step 1 and original sun image is normalized using mean filter, by gradation of image grade from 16 potential drops to 8
Position.
3. the solar activity object detection method according to claim 1 based on square net structure, it is characterised in that：Institute
Network in step 2 is stated to be defined as follows：
Wherein, [M, N] indicates original image size, is M rows, N row, [m, n] indicates the size of each grid image, INT functions
Judge whether number is shaping,WithRespectively indicate M and N open square root be integer as a result, floor
(M) and floor (N) indicates the integer no more than M and N respectively, and SQRT [floor (M)] and SQRT [floor (N)] are indicated respectively
The integer that can be opened quadratic power no more than M and N.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610374018.7A CN106067175B (en) | 2016-05-31 | 2016-05-31 | A kind of solar activity object detection method based on square net structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610374018.7A CN106067175B (en) | 2016-05-31 | 2016-05-31 | A kind of solar activity object detection method based on square net structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106067175A CN106067175A (en) | 2016-11-02 |
| CN106067175B true CN106067175B (en) | 2018-10-02 |
Family
ID=57420926
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610374018.7A Active CN106067175B (en) | 2016-05-31 | 2016-05-31 | A kind of solar activity object detection method based on square net structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106067175B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110211086B (en) * | 2018-02-28 | 2023-04-28 | 腾讯科技（深圳）有限公司 | Image segmentation method, device and storage medium |
| CN109753888A (en) * | 2018-12-17 | 2019-05-14 | 昆明理工大学 | A method of solar filament is chiral in astronomical image for identification |
| CN111127512B (en) * | 2019-11-29 | 2022-06-07 | 昆明理工大学 | Method for detecting and tracking solar activity area in full-sun image |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103455797A (en) * | 2013-09-07 | 2013-12-18 | 西安电子科技大学 | Detection and tracking method of moving small target in aerial shot video |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009009092A1 (en) * | 2007-07-10 | 2009-01-15 | Siemens Medical Solutions Usa, Inc. | System and method for detecting spherical and ellipsoidal objects using cutting planes |
-
2016
- 2016-05-31 CN CN201610374018.7A patent/CN106067175B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103455797A (en) * | 2013-09-07 | 2013-12-18 | 西安电子科技大学 | Detection and tracking method of moving small target in aerial shot video |
Non-Patent Citations (3)
| Title |
|---|
| Automatic Segmentation of Granules of the Solar Photosphere Using Morphological Reconstruction and Watershed Transform;Song Feng et al.;《2013 6th International Conference on Intelligent Networks and Intelligent Systems》;20131101;第300-303页 * |
| Automatic Solar Filament Segmentation and Characterization;Y. Yuan et al.;《Solar Physics》;20110831;第272卷;第101-117页 * |
| 一种提取目标图像最小外接矩形的快速算法;卢蓉 等;《计算机工程》;20101115;第36卷(第21期);第178-180页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106067175A (en) | 2016-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103440635B (en) | A kind of contrast limited adaptive histogram equalization method based on study | |
| Celik | Two-dimensional histogram equalization and contrast enhancement | |
| CN104376556B (en) | A kind of rock CT images Target Segmentation method | |
| CN102496023B (en) | Region of interest extraction method of pixel level | |
| CN104951799B (en) | A kind of SAR remote sensing image oil spilling detection recognition method | |
| CN106067175B (en) | A kind of solar activity object detection method based on square net structure | |
| Lomte et al. | Plant leaves image segmentation techniques: a review | |
| CN106529543B (en) | A kind of dynamic calculates the method and its system of polychrome grade binaryzation adaptive threshold | |
| CN103530644A (en) | Automatic identification method for weave structures of striped fabric | |
| CN105095907B (en) | A kind of heterogeneous fibre identification method of cotton based on RBF neural | |
| CN107133558A (en) | A kind of infrared pedestrian's conspicuousness detection method based on probability propagation | |
| CN109886267A (en) | A kind of soft image conspicuousness detection method based on optimal feature selection | |
| CN103020933A (en) | Multi-source image fusion method based on bionic visual mechanism | |
| CN108664839A (en) | A kind of image processing method and equipment | |
| CN105844213B (en) | Green fruit recognition method | |
| CN107154044A (en) | A kind of dividing method of Chinese meal food image | |
| CN109447111A (en) | A kind of remote sensing supervised classification method based on subclass training sample | |
| CN103218819A (en) | Automatic selection method for optimal homogenous region of ultrasound image and based on quad-tree decomposition | |
| CN104794726B (en) | A kind of underwater picture Parallel segmentation method and device | |
| CN109829896A (en) | The micro-calcification clusters automatic testing method of digital galactophore tomography X image based on multi-domain characteristics | |
| CN112528947A (en) | False hypha detection method and device by increasing direction dimension and storage medium | |
| CN105205485B (en) | Large scale image partitioning algorithm based on maximum variance algorithm between multiclass class | |
| CN112800968B (en) | HOG blocking-based feature histogram fusion method for identifying identity of pigs in drinking area | |
| CN109447997A (en) | A kind of underwater sonar image object dividing method of adaptive Cultural Algorithm | |
| Hsu et al. | Region-based image fusion with artificial neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |