CN104992183A - Method for automatic detection of substantial object in natural scene - Google Patents
Method for automatic detection of substantial object in natural scene Download PDFInfo
- Publication number
- CN104992183A CN104992183A CN201510377186.7A CN201510377186A CN104992183A CN 104992183 A CN104992183 A CN 104992183A CN 201510377186 A CN201510377186 A CN 201510377186A CN 104992183 A CN104992183 A CN 104992183A
- Authority
- CN
- China
- Prior art keywords
- pixel
- target
- fixation
- pixels
- area
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Abstract
The invention discloses a method for automatic detection of a significant object in natural scene. The method comprises the following steps: 1), performing significance detection on an object image to obtain a pixel significant graph; 2), ordering significant points in the pixel significant graph according to significance; 3), selecting the first N significant points as fixation points to form an area of fixation; 4), carrying out random sampling on pixels in the area of fixation, and performing equivalent pixel random sampling outside the area of fixation, wherein obtained pixels inside the area of fixation are taken as a positive sample, and pixels outside the area of fixation are taken as a negative sample; and 5), by use of a support vector machine training strategy, classifying all the pixels of the object image, and taking a pixel area classified as the positive sample as a first detection result; repeating the third step to the fifth step to obtain repeated detection results, and when the detection results are stable, recording the area; and repeating the second step to the fifth step until the image has no detection results. According to the invention, the visual sense of mankind is simulated through fixation points ordering and a nerve network model, and automatic detection of an object scene by a machine is realized.
Description
Technical field
The present invention relates to human visual simulation technical field, the automatic testing method of the well-marked target specifically in a kind of natural scene.
Background technology
Along with the development of infotech, computer vision has been widely used in the fields such as low-level feature detection and description, pattern-recognition, artificial intelligence reasoning and machine learning algorithm.But traditional computer vision methods is task-driven type normally, namely need to limit many conditions, and design corresponding algorithm according to actual task, lack versatility; Need to solve high dimensional nonlinear feature space, super large data volume to problems such as problem solving and process in real time, make its investigation and application face huge challenge.
Human visual system can efficiently, reliably work under various circumstances, and it has the following advantages: have the selectivity in the mechanism of concern, conspicuousness detection and visual processes related to this and purpose; Priori can be utilized from Low Level Vision process, make the bottom-up process of data-driven and top-down knowledge instruct mutual cooperation in visual processes; Upper and lower environment information all plays an important role the at all levels of visual processes, and can fully utilize the information of various mode in environment.But when human visual perception mechanism still not exclusively understands, still there is larger difficulty in the machine vision that how there is human vision characteristics, if the Vision Builder for Automated Inspection of simulating human vision can be built, bring important impact will inevitably to each practical application area of computer vision.
Summary of the invention
In view of this, the technical problem to be solved in the present invention is, there is provided a kind of can the automatic testing method of well-marked target in the natural scene of simulating human vision, by the behavior of simulating human active vision, do effectively to watch attentively fast to target scene, realize the automatic detection of machine to well-marked target in target scene.
Technical solution of the present invention is, provides the automatic testing method of the well-marked target in the natural scene of following steps, comprises following steps:
1) make conspicuousness by spectrum residual error method to target image to detect, obtain corresponding pixel saliency map, described pixel saliency map is consistent with the picture element position information of described target image;
2) to the significant point in described pixel saliency map, sort according to significance;
3) choose top n significant point as blinkpunkt, comprise the minimum rectangle scope of these blinkpunkts as watching area;
4) stochastic sampling is carried out to described watching area interior pixels, and the pixel stochastic sampling of equivalent is carried out to watching area outside; The watching area interior pixels that sampling obtains is as positive sample, and watching area external pixels is as negative sample;
5) utilize support vector machine ensembles Training strategy, training obtain multiple two classification SVM models, by whole pixels of target image described in these categories of model, the pixel region being divided into positive sample is done ballot integrated after, as the first testing result;
Choose a front N+M significant point as blinkpunkt, according to step 3) form watching area, then through step 4) and 5) obtain corresponding second testing result;
The relatively overlapping degree of the first testing result and the second testing result, overlapping degree greatly then shows the visually-perceptible intensity of target large; Overlapping degree is little, shows also not form the enough visually-perceptible intensity to target, continues to repeat said process, until reach enough visually-perceptible intensity, final testing result is the superposition of all testing results of said process;
After obtaining final testing result, in target image and pixel saliency map, this region is cleared, to the significant point in the pixel saliency map after renewal, according to significance minor sort again, repeat step 3), 4) and 5), obtain new testing result, until all target detection in target image are complete.
Adopt method of the present invention, compared with prior art, the present invention has the following advantages: carry out conspicuousness detection by spectrum residual error method, can form pixel saliency map fast; According to significance sorted pixels, can the high watching area of coarse localization significance; A small amount of pixel sampling is carried out to this intra-zone and outside simultaneously, form the training of positive and negative sample data and practice SVM (support vector machine) model, subsequently by this SVM category of model pixel, can obtain significance high, more accurate region is as the first testing result; And on the basis setting up the first testing result, can suitably expand the high watching area scope of significance, again form corresponding testing result through SVM study-classification, and compare with the first testing result, to judge whether fixation object district stablizes.The process that the present invention watches attentively according to human vision, by blinkpunkt sequence and SVM model, carrys out simulating human vision, realizes the automatic detection of machine to well-marked target in target scene.
As improvement, described spectrum residual error method refers to by supercomplex Fourier transform, red, green, blue in coloured image three components are participated in Fourier transform as hypercomplex three imaginary parts, only retains amplitude spectrum residual sum phase spectrum information, obtain pixel saliency map through inverse fourier transform.This is designed for and solves the problem that prior art only can process gray level image, effectively correspondingly improves the concrete steps of former spectrum residual error method for coloured image.
As improvement, described stochastic sampling for the Grad that has of pixel be greater than the average gradient value of its region.This is owing to being greater than former figure information entropy by the information entropy of gradient larger pixel generation in image, showing that high gradient pixel is representative to watched attentively target area, contribute to removal of images information redundancy.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the automatic testing method of well-marked target in natural scene of the present invention.
Embodiment
With regard to specific embodiment, the invention will be further described below, but the present invention is not restricted to these embodiments.
The present invention contain any make on marrow of the present invention and scope substitute, amendment, equivalent method and scheme.To have the present invention to make the public and understand thoroughly, in the following preferred embodiment of the present invention, describe concrete details in detail, and do not have the description of these details also can understand the present invention completely for a person skilled in the art.In addition, the needs in order to illustrate in the accompanying drawing of the present invention, completely accurately do not draw according to actual ratio, are explained at this.
As shown in Figure 1, the automatic testing method of the well-marked target in natural scene of the present invention, comprises following steps:
1) make conspicuousness by spectrum residual error method to target image to detect, obtain corresponding pixel saliency map, described pixel saliency map is consistent with the picture element position information of described target image;
2) to the significant point in described pixel saliency map, sort according to significance;
3) choose top n significant point as blinkpunkt, comprise the minimum rectangle scope of these blinkpunkts as watching area;
4) stochastic sampling is carried out to described watching area interior pixels, and the pixel stochastic sampling of equivalent is carried out to watching area outside; The watching area interior pixels that sampling obtains is as positive sample, and watching area external pixels is as negative sample;
5) utilize support vector machine ensembles Training strategy, training obtain multiple two classification SVM models, by whole pixels of target image described in these categories of model, the pixel region being divided into positive sample is done ballot integrated after, as the first testing result;
Choose a front N+M significant point as blinkpunkt, according to step 3) form watching area, then through step 4) and 5) obtain corresponding second testing result;
The relatively overlapping degree of the first testing result and the second testing result, overlapping degree greatly then shows the visually-perceptible intensity of target large; Overlapping degree is little, shows also not form the enough visually-perceptible intensity to target, continues to repeat said process, until reach enough visually-perceptible intensity, final testing result is the superposition of all testing results of said process;
After obtaining final testing result, in target image and pixel saliency map, this region is cleared, to the significant point in the pixel saliency map after renewal, according to significance minor sort again, repeat step 3), 4) and 5), obtain new testing result, until all target detection in target image are complete.
Natural scene is equivalent to the scene that human vision is watched attentively, no matter scene size, the scope of imaging on the retina constant, and thus natural scene is also like this in machine in machine vision.
Make conspicuousness by spectrum residual error method to target image to detect, following steps can be adopted to implement: treat perceptual image I (x) (x represents pixel coordinate vector) herein for given, first two dimensional discrete Fourier transform F [I (x)] is carried out to it, image is changed to frequency domain by transform of spatial domain, obtains amplitude A (f) and phase place P (f) information:
A(f)=|F[I(x)]| (1)
Then amplitude is taken the logarithm, obtains log and compose L (f):
L(f)=log(A(f)) (3)
In formula, F represents two dimensional discrete Fourier transform, | .| represents amplitude computing,
represent phase bit arithmetic.Because log curve meets local linear condition, so with local average wave filter h
nf () is smoothing to it, obtain the general shape of log spectrum:
V(f)=L(f)*h
n(f) (4)
Wherein h
n(f) be a n × n matrix (the present embodiment experiment in n=3.), be defined as follows:
Spectrum residual error R (f) is then the description to the Sudden change region in image:
R(f)=L(f)-V(f) (6)
By inverse Fourier transform, saliency map picture can be obtained in spatial domain.
S(x)=|F
-1[exp{R(f)+jP(f)}]|
2(7)
On saliency map, the value of often represents the significance of this position.Considering the local group effect of human eye vision, in order to eliminate the isolated significant point of minority, obtaining better visual effect, can carry out once level and smooth with average filter again after obtaining S (x), obtain final saliency map Z (x).
Z(x)=S(x)*h
n(f) (8)
In Fig. 1, relate to training data, disaggregated model, result etc. and be the corresponding implementation process of employing support vector machine (SVM) Training strategy.Specific implementation process is as follows:
If comprise the training set of l sample
for input vector, y
k{-1 ,+1} is positive and negative classification logotype to ∈.First SVM uses training set learning model building, and object finds optimal separating hyper plane at feature space, test data as far as possible correctly classified.Consider generalized case, when training set is Nonlinear separability, first select a gaussian radial basis function kernel function
K(x,x
i)=exp{-q||x-x
i||
2} (9)
By training set data x
ibe mapped in a High-dimensional Linear feature space and construct optimal separating hyper plane.Wherein q is Radial basis kernel function parameter, then the discriminant function of sorter is
Training process is known
with under the condition such as q, Quadratic Programming Solution method is utilized to obtain b in (10) formula
*, α
i *with support vector (SV) as training the SVM model obtained; Test process is then utilize this SVM model, the data x of the unknown is substituted into (10) formula, obtains its prediction classification.
The dimension disaster problem that SVM utilizes kernel function skill to avoid traditional learning algorithm to face.Structure based principle of minimization risk, its classification performance is only determined by a small amount of support vector (SV), the Generalization Capability had.In practical problems, be conducive to utilizing priori to select a small amount of sample, carry out structural classification device through SVM study.Which overcome traditional learning algorithm based on empirical risk minimization principle, when sample number trends towards infinity, performance just has the defect of theoretic guarantee; By solving quadratic programming problem, traditional neural network algorithm can be avoided to build the empirical of network and be easily absorbed in the shortcomings such as local minimizers number; Be applicable to detection of complex, be difficult to the image object of quantitative description.
During SVM practical application, for dissimilar image, need to adjust the classification performance that wherein some key parameters just can obtain.In order to reduce the adverse effect that svm classifier Model Parameter is arranged, the training set that the positive and negative composition of sample of this method multi collect is slightly different, and adopt differentiated Parameter Parallel that multiple slightly different SVM model is set.Object carries out parallel training and test to the svm classifier model that parameter is slightly different respectively with the different training sets of slightly disturbance, finally carries out ballot method to test result integrated.Such strategy can increase the robustness of disaggregated model, greatly reduces SVM model parameter and arranges the improper adverse effect caused.
Described spectrum residual error method can also pass through supercomplex Fourier transform, red, green, blue in coloured image three components are participated in Fourier transform as hypercomplex three imaginary parts, only retain amplitude spectrum residual sum phase spectrum information, obtain pixel saliency map through inverse fourier transform.This is designed for and solves the problem that prior art only can process black white image identification, effectively correspondingly improves the concrete steps of tradition spectrum residual error method for coloured image.
Supercomplex is made up of four parts, is expressed as
q=a+bi+cj+dk (11)
Wherein a, b, c, d are real numbers, i, j, k Dou Shi imaginary unit, and have following character: i
2=j
2=k
2=ijk=-1, ij=-ji=k, ki=-ik=j, jk=-kj=i.
The RGB model of coloured image can be described as the pure supercomplex not having real part:
f=R(m,n)i+G(m,n)j+B(m,n)k (12)
Wherein R (m, n), G (m, n), B (m, n) represent image RGB three components respectively.If q=f, then a=0, b=R (m, n), c=G (m, n), d=B (m, n).Supercomplex Fourier transform can be carried out according to formula (13) to the colour phasor constructed:
F
R(v,u)=(real(fft2(a))+μ·imag(fft2(a)))+
i(real(fft2(b))+μ·imag(fft2(b)))+ (13)
j(real(fft2(c))+μ·imag(fft2(c)))+
k(real(fft(d))+μ·imag(fft2(d)))
Wherein, fft2 () represents conventional two-dimensional Fourier transform, and real part is got in real () expression, and imaginary part is got in imag () expression.
for the empty vector of unit.Herein, only F need be got
ramplitude spectrum residual error R (f) of (v, u) and phase spectrum p (f):
R(f)=log|F
R(v,u)|-h*log|F
R(v,u)| (14)
Wherein, h is local average operator.Order:
A=e
R(f)+jP(f)(16)
Utilize conventional two-dimensional inverse fast Fourier transform (ifft2) to combine and can obtain supercomplex inverse Fourier transform, such as formula (17):
F
-R(v,u)=(real(ifft2(A))+μ·imag(ifft2(A)))+
i(real(ifft2(B))+μ·imag(ifft2(B)))+ (17)
j(real(ifft2(C))+μ·imag(ifft2(C)))+
k(real(ifft2(D))+μ·imag(ifft2(D)))
Wherein, B=fft2 (b), C=fft2 (c), D=fft2 (d).
Real (F
-R(v, u)) be the remarkable figure tried to achieve.Because the globality of colour element before and after data processing obtains maintenance, thus avoid the color distortion that conversion or exchange due to vector component cause.
Described stochastic sampling for the Grad that has of pixel be greater than the average gradient value of its region.Show the research of image information entropy, the information entropy having the partial pixel of higher gradient to produce in image is greater than the information entropy that source images entire pixels is formed.This is the phenomenon because image information redundancy causes.Given this phenomenon, in order to obtain the richest quantity of information, the representational pixel samples in watching area, sampling should be carried out for high gradient pixel.In order to avoid a large amount of calculating, a kind of effective countermeasure is: the pixel being only greater than this zone leveling Grad for watching area inner gradient value carries out stochastic sampling.And the sampling of watching area outside is still taked entire pixels stochastic sampling mode.
Below only just preferred embodiment of the present invention is described, but can not be interpreted as it is limitations on claims.The present invention is not only confined to above embodiment, and its concrete structure allows to change.In a word, all various changes done in the protection domain of independent claims of the present invention are all in protection scope of the present invention.
Claims (3)
1. an automatic testing method for the well-marked target in natural scene, is characterized in that: comprise the following steps:
1) make conspicuousness by spectrum residual error method to target image to detect, obtain corresponding pixel saliency map, described pixel saliency map is consistent with the picture element position information of described target image;
2) to the significant point in described pixel saliency map, sort according to significance;
3) choose top n significant point as blinkpunkt, comprise the minimum rectangle scope of these blinkpunkts as watching area;
4) stochastic sampling is carried out to described watching area interior pixels, and the pixel stochastic sampling of equivalent is carried out to watching area outside; The watching area interior pixels that sampling obtains is as positive sample, and watching area external pixels is as negative sample;
5) utilize support vector machine Training strategy, training obtains the SVM model of one two classification, by whole pixels of target image described in this category of model, will be divided into the pixel region of positive sample as the first testing result;
Choose a front N+M significant point as blinkpunkt, according to step 3) form watching area, then through step 4) and 5) obtain corresponding second testing result;
The relatively overlapping degree of the first testing result and the second testing result, overlapping degree greatly then shows the visually-perceptible intensity of target large; Overlapping degree is little, shows also not form the enough visually-perceptible intensity to target, continues to repeat said process, until reach enough visually-perceptible intensity, final testing result is the superposition of all testing results of said process;
After obtaining final testing result, in target image and pixel saliency map, this region is cleared, to the significant point in the pixel saliency map after renewal, according to significance minor sort again, repeat step 3), 4) and 5), obtain new segmentation result, until all target detection in target image are complete.
2. the automatic testing method of the well-marked target in natural scene according to claim 1 and 2, it is characterized in that: described spectrum residual error method refers to by supercomplex Fourier transform, red, green, blue in coloured image three components are participated in Fourier transform as hypercomplex three imaginary parts, only retain amplitude spectrum residual sum phase spectrum information, obtain pixel saliency map through inverse fourier transform.
3. the automatic testing method of the well-marked target in natural scene according to claim 1 and 2, is characterized in that: described stochastic sampling for the Grad that has of pixel be greater than the average gradient value of its region.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510377186.7A CN104992183B (en) | 2015-06-25 | 2015-06-25 | The automatic testing method of well-marked target in natural scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510377186.7A CN104992183B (en) | 2015-06-25 | 2015-06-25 | The automatic testing method of well-marked target in natural scene |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104992183A true CN104992183A (en) | 2015-10-21 |
CN104992183B CN104992183B (en) | 2018-08-28 |
Family
ID=54303996
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510377186.7A Active CN104992183B (en) | 2015-06-25 | 2015-06-25 | The automatic testing method of well-marked target in natural scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104992183B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105956592A (en) * | 2016-05-10 | 2016-09-21 | 西北工业大学 | Aircraft target detection method based on image significance and SVM |
CN106815604A (en) * | 2017-01-16 | 2017-06-09 | 大连理工大学 | Method for viewing points detecting based on fusion of multi-layer information |
CN107992875A (en) * | 2017-12-25 | 2018-05-04 | 北京航空航天大学 | A kind of well-marked target detection method based on image bandpass filtering |
CN108897786A (en) * | 2018-06-08 | 2018-11-27 | Oppo广东移动通信有限公司 | Recommended method, device, storage medium and the mobile terminal of application program |
CN109190473A (en) * | 2018-07-29 | 2019-01-11 | 国网上海市电力公司 | The application of a kind of " machine vision understanding " in remote monitoriong of electric power |
CN110415240A (en) * | 2019-08-01 | 2019-11-05 | 国信优易数据有限公司 | Sample image generation method and device, circuit board defect detection method and device |
CN111481166A (en) * | 2017-05-04 | 2020-08-04 | 深圳硅基智能科技有限公司 | Automatic identification system based on eye ground screening |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101980248A (en) * | 2010-11-09 | 2011-02-23 | 西安电子科技大学 | Improved visual attention model-based method of natural scene object detection |
US7940985B2 (en) * | 2007-06-06 | 2011-05-10 | Microsoft Corporation | Salient object detection |
CN102945378A (en) * | 2012-10-23 | 2013-02-27 | 西北工业大学 | Method for detecting potential target regions of remote sensing image on basis of monitoring method |
-
2015
- 2015-06-25 CN CN201510377186.7A patent/CN104992183B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7940985B2 (en) * | 2007-06-06 | 2011-05-10 | Microsoft Corporation | Salient object detection |
CN101980248A (en) * | 2010-11-09 | 2011-02-23 | 西安电子科技大学 | Improved visual attention model-based method of natural scene object detection |
CN102945378A (en) * | 2012-10-23 | 2013-02-27 | 西北工业大学 | Method for detecting potential target regions of remote sensing image on basis of monitoring method |
Non-Patent Citations (3)
Title |
---|
XIAODI HOU: "Saliency Detection: A Spectral Residual Approach", 《COMPUTER VISION AND PATTERN RECOGNITION(CVPR),2007IEEE CONFERENCE ON》 * |
侯庆岑: "模拟人类视觉的自动图像分割技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
潘晨 等: "基于空间和时间差别采样的色彩图像分割", 《计算机工程》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105956592A (en) * | 2016-05-10 | 2016-09-21 | 西北工业大学 | Aircraft target detection method based on image significance and SVM |
CN105956592B (en) * | 2016-05-10 | 2019-03-29 | 西北工业大学 | A kind of Aircraft Targets detection method based on saliency and SVM |
CN106815604A (en) * | 2017-01-16 | 2017-06-09 | 大连理工大学 | Method for viewing points detecting based on fusion of multi-layer information |
CN106815604B (en) * | 2017-01-16 | 2019-09-27 | 大连理工大学 | Method for viewing points detecting based on fusion of multi-layer information |
CN111481166A (en) * | 2017-05-04 | 2020-08-04 | 深圳硅基智能科技有限公司 | Automatic identification system based on eye ground screening |
CN111481166B (en) * | 2017-05-04 | 2021-11-26 | 深圳硅基智能科技有限公司 | Automatic identification system based on eye ground screening |
CN107992875A (en) * | 2017-12-25 | 2018-05-04 | 北京航空航天大学 | A kind of well-marked target detection method based on image bandpass filtering |
CN107992875B (en) * | 2017-12-25 | 2018-10-26 | 北京航空航天大学 | A kind of well-marked target detection method based on image bandpass filtering |
CN108897786A (en) * | 2018-06-08 | 2018-11-27 | Oppo广东移动通信有限公司 | Recommended method, device, storage medium and the mobile terminal of application program |
CN108897786B (en) * | 2018-06-08 | 2021-06-08 | Oppo广东移动通信有限公司 | Recommendation method and device of application program, storage medium and mobile terminal |
CN109190473A (en) * | 2018-07-29 | 2019-01-11 | 国网上海市电力公司 | The application of a kind of " machine vision understanding " in remote monitoriong of electric power |
CN110415240A (en) * | 2019-08-01 | 2019-11-05 | 国信优易数据有限公司 | Sample image generation method and device, circuit board defect detection method and device |
Also Published As
Publication number | Publication date |
---|---|
CN104992183B (en) | 2018-08-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104992183A (en) | Method for automatic detection of substantial object in natural scene | |
CN110210486B (en) | Sketch annotation information-based generation countermeasure transfer learning method | |
CN104966085B (en) | A kind of remote sensing images region of interest area detecting method based on the fusion of more notable features | |
CN104166859B (en) | Based on SSAE and FSALS SVM Classification of Polarimetric SAR Image | |
CN113192040A (en) | Fabric flaw detection method based on YOLO v4 improved algorithm | |
CN107545263B (en) | Object detection method and device | |
CN107491734B (en) | Semi-supervised polarimetric SAR image classification method based on multi-core fusion and space Wishart LapSVM | |
CN107194872A (en) | Remote sensed image super-resolution reconstruction method based on perception of content deep learning network | |
CN111079685A (en) | 3D target detection method | |
CN108108751A (en) | A kind of scene recognition method based on convolution multiple features and depth random forest | |
CN112164054A (en) | Knowledge distillation-based image target detection method and detector and training method thereof | |
CN104282008A (en) | Method for performing texture segmentation on image and device thereof | |
CN112613350A (en) | High-resolution optical remote sensing image airplane target detection method based on deep neural network | |
CN105894013A (en) | Method for classifying polarized SAR image based on CNN and SMM | |
CN107392863A (en) | SAR image change detection based on affine matrix fusion Spectral Clustering | |
CN114463637A (en) | Winter wheat remote sensing identification analysis method and system based on deep learning | |
CN104933691A (en) | Image fusion method based on phase spectrum visual saliency detection | |
Müller et al. | Simulating optical properties to access novel metrological parameter ranges and the impact of different model approximations | |
CN102737232B (en) | Cleavage cell recognition method | |
CN106446965A (en) | Spacecraft visible light image classification method | |
CN109711420A (en) | The detection and recognition methods of alveolar hydalid target based on human visual attention mechanism | |
CN104933435B (en) | Machine vision construction method based on simulation human vision | |
CN105005788A (en) | Target perception method based on emulation of human low level vision | |
CN105023016B (en) | Target apperception method based on compressed sensing classification | |
Veeravasarapu et al. | Model-driven simulations for computer vision |
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 |