CN106485239A - One kind is using one-class support vector machines detection river mesh calibration method - Google Patents
One kind is using one-class support vector machines detection river mesh calibration method Download PDFInfo
- Publication number
- CN106485239A CN106485239A CN201610944422.3A CN201610944422A CN106485239A CN 106485239 A CN106485239 A CN 106485239A CN 201610944422 A CN201610944422 A CN 201610944422A CN 106485239 A CN106485239 A CN 106485239A
- Authority
- CN
- China
- Prior art keywords
- river
- detection
- target
- support vector
- vector machines
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 27
- 230000008569 process Effects 0.000 claims abstract description 20
- 230000003595 spectral effect Effects 0.000 claims abstract description 15
- 239000000284 extract Substances 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims abstract description 13
- 238000004513 sizing Methods 0.000 claims abstract description 11
- 238000003709 image segmentation Methods 0.000 claims abstract description 8
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 238000004458 analytical method Methods 0.000 claims description 12
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 10
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 claims description 7
- 238000001228 spectrum Methods 0.000 claims description 7
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000004069 differentiation Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 2
- 238000012360 testing method Methods 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001568 sexual effect Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000282320 Panthera leo Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000007429 general method Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000000227 grinding Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/182—Network patterns, e.g. roads or rivers
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
In order to solve the problems, such as to be learnt by training sample in prior art and testing result is not accurate enough, the present invention provides one kind using one-class support vector machines detection river mesh calibration method, through coarse sizing process after spectral signature in extraction remote sensing images, based on spectral signature, extract river candidate region, recycle the characteristic vector that spectral value generates that one-class support vector machines are trained.Result for coarse sizing process is finely detected again, so that river candidate region is carried out with image segmentation, generates shape facility, finally according to shape index threshold value, realizes target detection.The present invention only needs a class training sample using one-class support vector machines method, improves detection efficiency；Improve the accuracy of detection by coarse sizing and fine two links of detection simultaneously.
Description
Technical field
The present invention relates to image procossing and computer vision field, more particularly, to one kind utilize one-class support vector machines method
River mesh calibration method in detection remote sensing images.
Background technology
Object detecting and tracking is one of popular research direction of image procossing and computer vision field, militarily
Imaging Guidance, follow the tracks of military target, and the aspect such as the safety monitoring of civil aspect, intelligent human-machine interaction all has important grinding
Study carefully value.
In existing detection method, the difficult point of target detection research is the Efficient Characterization to target, and a variety of causes
The object matching problem that the target scale that causes, the anglec of rotation, illumination etc. change and causes.Target be characterized in certain journey
Degree determines matching algorithm, including：Using local configuration character representation target, extract the profile of image object first, and pass through
Automatically selected threshold filters off noise edge, obtains significance profile, reduces profile hop count so as to effective, reduces subsequent characteristics and carries
Take and whole detection process Space-time Complexity.In order to overcome yardstick, rotation of target etc. to change, it is special that Lowe proposes Scale invariant
Levy conversion (SIFT), by calculating multiple dimensioned difference of Gaussian image the method finding Local modulus maxima, obtain in certain model
Enclose interior yardstick, invariable rotary feature.
In target detection, in addition to the feature of target itself, can also be characterized using context constraint, realize image mesh
Target effective detection.In Remote Sensing Target detection, because picture size is larger and complicated, more in order to avoid producing
False-alarm, needs to take certain measure to remove false-alarm targets as far as possible.The general method adopting is first to filter out time by simple feature
Select target area, then again candidate region is finely detected.Have again, the detection for remote sensing images Aircraft Targets is asked
Topic, is first gone out candidate target window using tandem type detection of classifier, then with Hough forest algorithm, candidate window is carried out secondary
Judge, filter false-alarm, improve detection efficiency, save operation time.
Conventional sorting technique such as K- nearest neighbor classifier, bayes classification method all have complex the lacking of process
Point.
If K- arest neighbors is the most straightaway one kind in classifier algorithm, calculate test sample to each training sample away from
From taking wherein minimum K, and carry out voting according to the labelling of this K training sample and obtain the labelling of test sample.Algorithm
Clear thinking is simple, but excessive for mass data amount of calculation, and each training sample has a distance must measure, and expends
Plenty of time.
Bayes classification method needs to complete to calculate using multiple stages：Preparation stage first stage, this
The task in stage is to do necessary preparation for Naive Bayes Classification, and groundwork is to determine characteristic attribute as the case may be,
And each characteristic attribute is suitably divided, then by manually classifying to a part of item to be sorted, form training sample
Set.The input in this stage is all data to be sorted, and output is characteristic attribute and training sample.This stage is whole Piao
The stage being accomplished manually uniquely is needed, its quality will have a major impact to whole process, the matter of grader in plain Bayes's classification
Amount is largely divided by characteristic attribute, characteristic attribute and training sample quality determines.
In the second stage classifier training stage, the task in this stage is exactly to generate grader, and groundwork is meter
Calculate the frequency of occurrences in training sample for each classification and each characteristic attribute divides the conditional probability to each classification and estimates, and
By result record.Its input is characteristic attribute and training sample, and output is grader.This stage is the mechanical sexual stage, according to
Formula previously discussed can be calculated by Automatic Program and complete.
Application stage phase III.The task in this stage is to treat sorting item using grader to be classified, its
Input is grader and item to be sorted, and output is the mapping relations of item to be sorted and classification.This stage is also the mechanical sexual stage,
Completed by program.
River is the important goal in remote sensing images, and river target detection is all widely used at military and civil aspect.
In prior art, target detection link, image object detection often utilize various features such as color (spectrum) feature, shape
Architectural feature, textural characteristics, contextual feature, SIFT feature etc..In realistic objective detection application, the principle of feature selection is
Effectively target should be able to be characterized, for different targets, feature selection result is also different.
Spectral signature is the gray value of image pixel, represents the spectral reflection characteristic of ground target, through frequently as atural object class
The foundation of not other division, but the geometry feature of target to be detected can not be reflected.The image that original image produces after over-segmentation
Speckle has shape and structure feature, based on one covariance matrix of vectorial structure of the pixel coordinate composition in image patch, enters
One step can extract the features such as the length-width ratio of this image patch, shape index, density, principal direction.Feature based on image segmentation
Extract, affected by dividing method and result very big, therefore there is higher uncertainty.And the method for expressing of texture is a lot,
It is most commonly used that the textural characteristics description based on gray level co-occurrence matrixes.The joint of the pixel with two positions for the gray level co-occurrence matrixes is general
Defining, it not only reflects the distribution character of brightness to rate density, and also reflection has same brightness or between the pixel of brightness
Position distribution characteristic, be the second-order statisticses feature of relevant image brightness change, be the basis defining one group of textural characteristics.Texture
, for general target floor target, distinction is often strong for feature, conventional textural characteristics have homogeneity, contrast, energy,
Entropy etc..
In a word, in existing technology：1. the feature adopting for remote sensing images river object detection method has color (light
Spectrum) feature, shape and structure feature, textural characteristics, contextual feature, SIFT feature etc., algorithm complexity, the result of detection is not accurate
Really.
2. detect specific target classification in object detection task it is desirable to from background, and the class comprising in background
Other quantity does not know, and typically the classification in background cannot be learnt automatically.
Content of the invention
In order to solve the problems, such as to be learnt by training sample in prior art and testing result is not accurate enough, this
Invention provides one kind using one-class support vector machines detection river mesh calibration method, and it has：Only need a class training sample, adopt
With single class class vector machine method, overcome the difficulty of the other samples selection of non-target class in target detection, and make target detection mistake
The study stage in journey simplifies, and improves detection efficiency；Improve detection by coarse sizing and two links of fine detection simultaneously
Accuracy.
Solve problem of the present invention be employed technical scheme comprise that：Using following steps：
A. select wave band：In remote sensing images choose band combination carry out atural object distinguish and river extract so as to water body,
Vegetation and other ground objects are identified；
B. feature analysiss and selection：The image selecting A link to process, carries out Spectral Characteristics Analysis：Spectrum analyses adopt
Spectral signature be：The water body target such as river, in remote sensing images, distinguishes river target with gray value；
C. coarse sizing process：Based on spectral signature, extract river candidate region：Select some target samples, extract each
The spectral value of pixel as characteristic of division, using formula xi=(ri, gi, bi) generate characteristic vector, using the characteristic vector generating
One-class support vector machines are trained, during training, adopt RBF kernel functionIt is trained,
And determine disaggregated model by 10 folding cross validations, single class classification is carried out to whole remote sensing image, obtains water body classification extraction knot
Really；Wherein, riCorresponding red component；giCorresponding green component；biCorresponding blue component；||xi-xj| | any two in representation space
Point xiAnd xjBetween Euclidean distance；γ is nuclear parameter.
D. fine detection process：For the result in C link, river candidate region is carried out with image segmentation, generate shape
Feature：On the basis of C link coarse sizing link acquired results, using image Segmentation Technology, threshold parameter is set, will be greater than threshold
Value and adjacent pixel merges, generates different size of target classification connected region, then by the picture in each connected region
Element is calculated shape facility indexAnd set area threshold and carry out removing of small regions, finally will remove little face domain
Image be merged into background area；Wherein, the long e in border：The number of edge pixel, the border a length of 1 of a pixel；Area A：Group
Become the pixel sum of this object, the length of one of pixel edge is set to 1.
E. target detection：Set shape index threshold value, and the shape facility index being drawn according to D link determines river mesh
Mark, realizes target detection.
The fine detection process of described D link：Using threshold segmentation method, with the value of area A as threshold value, extract target
Candidate region, generates shape facility S simultaneously.
Further, described waveband selection link, selects 4,3,2 wave bands of remote sensing images, gives red, green, blue respectively
Color, carries out atural object differentiation.
Further, described feature analysiss and selection link gray value are 10～20.
Further, in described fine detection process, threshold parameter during image segmentation is：Th=10；Area threshold is：
AT=50.
Further, the shape index threshold value in described target detection link is ST=2.5.
The invention has the beneficial effects as follows：The present invention only needs a class training sample using one-class support vector machines method, gram
Take the difficulty of the other samples selection of non-target class in target detection, and the study stage during making target detection has simplified, and improves
Detection efficiency；Improve the accuracy of detection by coarse sizing and fine two links of detection simultaneously.
Brief description
Fig. 1 is remote sensing images schematic diagram.
Fig. 2 is the river candidate region schematic diagram being determined using one-class support vector machines method.
Fig. 3 is the river candidate region schematic diagram being determined using arest neighbors sorting technique.
Fig. 4 is the river candidate region schematic diagram being determined using bayes classification method.
Fig. 5 removes the zonule schematic diagram less than 50 pixels.
Fig. 6 shape facility index is more than matching result schematic diagram when 2.5.
Specific embodiment
As shown in Fig. 1～2, one kind detects river mesh calibration method using one-class support vector machines, using following steps：
A. the waveband selection of remote sensing images, chooses the band combination being suitable for that atural object is distinguished and river extracts：Select remote sensing figure
The 4 of picture, 3,2 wave bands, respectively give red, green, blue color, these wave bands be combined into false color image, its cartographic feature enrich, fresh
Bright, level is good, can preferably water body, vegetation etc. be identified.
B. feature analysiss and selection：The image selecting A link to process, carries out Spectral Characteristics Analysis：Spectrum analyses adopt
Spectral signature be the water body target such as river a marked feature：In remote sensing images, river target has relatively low gray scale
It is worth for 10～20.
C. coarse sizing process：Based on the spectrum analyses of gray feature in B link, extract river candidate region：First, select
As target class very originally, the spectral value extracting each pixel is as characteristic of division x for 1000 pixelsi=(ri, gi, bi), generate
1000 characteristic vectors xi, wherein, riFor red component, giFor green component, biFor blue component.Using these characteristic vectors
One-class support vector machines are trained.Given l training sample, one-class support vector machines duty Optimization, that is,
And meet (wTφ(xi)+b)≥1-ξi, ξi≥0 (2).
Wherein, C>0 is penalty coefficient, φ (xi) it is by vector xiIt is mapped to the function of higher dimensional space；W is in feature space
The normal vector of classifying face, wTIt is the transposition of w, b is the intercept of w, ξiIt is relaxation factor.
General, K (xi, xj)≡φ(xi)Tφ(xi) for the kernel function in support vector machine.
In the training process, characteristic vector and category label are combined, as training |input paramete, that is,（xi,
yi）, wherein yiIt is category label, in this sorting technique, yi=1, that is, classification is all positive class.RBF kernel function is selected in trainingWherein, | | xi-xj| | any two points x in representation spaceiAnd xjBetween Euclidean distance, and lead to
Cross 10 folding cross validations and obtain sorting parameter.10 folding cross validation methods are randomly this 1000 training samples to be divided into 10
Individual equal subset, alternately selects wherein 9 subsets as training, in addition 1 subset is as test, so, each subset
Test data will be used as, the ratio according to correct classification samples calculates the precision of cross validation.Convert different ginsengs
Array is closed, i.e. nuclear parameter γ-value in penalty coefficient C and kernel function, selects the combination obtaining optimal cross validation precision as list
Class support vector machines disaggregated model parameter, w, b also determine therewith simultaneously.Using the disaggregated model determining, to whole remote sensing image
Carry out single class classification, obtain water body classification extraction result.
One-class support vector machines only need to the other sample of river target class and are learnt, the river mesh required for just obtaining
Mark category classification result, decreases the task of training stage.And adopt multi-class classification method extract simple target classification, need by
Image division becomes multiple atural object classifications, the workload greatly increasing.
In the present embodiment, if Fig. 3, Fig. 4 are the river being obtained using arest neighbors classification and bayes classification method respectively
Objective extraction result.Using bayes classification method, need for experimental image to be divided into impermeable surface, forest land, meadow and water
Four classifications of body, and it is respectively each classification selection training sample, increased workload and the uncertainty in study stage, make total
Body classification difficulty increases.Compared with the one-class support vector machines method using a class training sample in Fig. 2, classification results very phase
Closely, and the training of one-class support vector machines method is more prone to.
D. fine detection process：River candidate region is carried out with image segmentation, generates shape facility：In C link coarse sizing
On the basis of link acquired results, using carrying out image threshold segmentation method, object candidate area is extracted, generate shape simultaneously
Feature.Due to only white and black picture element in the image of river candidate region, threshold parameter Th=10 is set, will be greater than threshold value and
Adjacent pixel merges, and generates different size of target classification connected region, then by the pixel meter in each connected region
Calculation obtains shape facility, computational methods such as formula（3）.
According to the target inspection requirements of actual river, remove zonule candidate target, zonule is carried out with area threshold AT=50
Remove, will area features A<50 candidate target removes, and is merged into background area, result is as shown in Figure 5.
Based on the actual application requirements, river target should have certain area, and associated shape feature description is as follows.
The long e in border：The number of edge pixel, the border a length of 1 of a pixel.
Area A：Form the pixel sum of this object, the length of one of pixel edge is set to 1.
Shape facility index：
The square root of the area divided by 4 times for the border length, S is the relation on description object border and area, and border is longer, and value is got over
Greatly.
E. river target is determined according to shape facility index, realize target detection：Shape information is that remote sensing image is visually sentenced
One of other very important factor, for width remote sensing images, by segmented extraction shape information, obtains one with many
The attribute of the shape facility of each image object and river target is carried out in this set by the image object set of individual attribute
Coupling, filters out river target.River target shape feature mainly has larger shape facility index, selects numerical value 2.5
Judged as threshold value.
On the basis of D link removes zonule target, according to the candidate target shape facility index of segmentation generation, by shape
Shape characteristic index threshold value is set to ST=2.5, carries out river target characteristic coupling.Shape facility index when candidate target region
Value S>When 2.5, retain as object matching result, if shape facility exponential quantity S≤2.5, as background removal, detection knot
Fruit is as shown in Figure 6.
During zonule object removal and characteristic matching, set administrative characteristic threshold value should be according to actual needs
Adjustment, different application requirements has different parameters.
The present invention is detected to river target using one-class support vector machines, and flow process is simple, and amount of calculation greatly reduces,
Accelerate work efficiency, be worth of widely use.
Claims (6)
1. one kind, using one-class support vector machines detection river mesh calibration method, is characterized in that：Comprise the steps of：
A. select wave band：Choose band combination and carry out atural object differentiation and river extraction, so as to water body, vegetation in remote sensing images
And other ground objects are identified；
B. feature analysiss and selection：The image selecting A link to process, carries out Spectral Characteristics Analysis：The light that spectrum analyses adopt
Spectrum signature is：The water body target such as river, in remote sensing images, distinguishes river target with gray value；
C. coarse sizing process：Based on spectral signature, extract river candidate region：Select some target samples, extract each pixel
Spectral value as characteristic of division, using formula xi=(ri, gi, bi) generate characteristic vector, using the characteristic vector generating to list
Class support vector machines carry out duty Optimization, adopt RBF kernel function in trainingInstructed
Practice, and determine disaggregated model by 10 folding cross validations, single class classification is carried out to whole remote sensing image, obtains water body classification extraction
Result；
Wherein, riFor red component, giFor green component, biFor blue component；
Wherein, | | xi-xj| | any two points x in representation spaceiAnd xjBetween Euclidean distance；γ is nuclear parameter；
D. fine detection process：For the result in C link, river candidate region is carried out with image segmentation, generate shape facility
Index：On the basis of C link coarse sizing link acquired results, using carrying out image threshold segmentation technology, threshold parameter is set, will be big
In threshold value and adjacent pixel merges, generate different size of target classification connected region, then by each connected region
Pixel be calculated shape facility indexAnd set area threshold and carry out removing of small regions, finally will remove little
The image in face domain is merged into background area；
Wherein, the long e in border：The number of edge pixel, the border a length of 1 of a pixel；
Area A：Form the pixel sum of this object, the length of one of pixel edge is set to 1；
E. target detection：Set shape index threshold value, and the shape facility index being drawn according to D link determines river target, real
Existing target detection.
2. utilization one-class support vector machines detection river according to claim 1 mesh calibration method, is characterized in that：Described
Waveband selection link, selects 4,3,2 wave bands of remote sensing images, gives red, green, blue color respectively and carries out atural object differentiation.
3. utilization one-class support vector machines detection river according to claim 1 mesh calibration method, is characterized in that：Described
Feature analysiss are 10～20 with selection link gray value.
4. utilization one-class support vector machines detection river according to claim 1 mesh calibration method, is characterized in that：Described
In fine detection process, threshold parameter during image segmentation is：Th=10；Area threshold is：AT=50.
5. utilization one-class support vector machines detection river according to claim 1 mesh calibration method, is characterized in that：Described
Shape index threshold value in target detection link is ST=2.5.
6. utilization one-class support vector machines detection river according to claim 1 mesh calibration method, is characterized in that：Described
In C link, the process of duty Optimization is：Given l training sample, then have：
And meet (wTφ(xi)+b)≥1-ξi, ξi≥0
Wherein, C>0 is penalty coefficient, φ (xi) it is the function that vector x i is mapped to higher dimensional space；W is classification in feature space
The normal vector in face, wTIt is the transposition of w, b is the intercept of w, ξiIt is relaxation factor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610944422.3A CN106485239A (en) | 2016-11-02 | 2016-11-02 | One kind is using one-class support vector machines detection river mesh calibration method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610944422.3A CN106485239A (en) | 2016-11-02 | 2016-11-02 | One kind is using one-class support vector machines detection river mesh calibration method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106485239A true CN106485239A (en) | 2017-03-08 |
Family
ID=58271342
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610944422.3A Pending CN106485239A (en) | 2016-11-02 | 2016-11-02 | One kind is using one-class support vector machines detection river mesh calibration method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106485239A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334909A (en) * | 2018-03-09 | 2018-07-27 | 南京天数信息科技有限公司 | Cervical carcinoma TCT digital slices data analysing methods based on ResNet |
| CN108710862A (en) * | 2018-05-24 | 2018-10-26 | 河海大学 | A kind of high-resolution remote sensing image Clean water withdraw method |
| CN108920998A (en) * | 2018-04-14 | 2018-11-30 | 海南大学 | Single category based on Pasteur's coefficient pours down remote sensing image target area extracting method |
| CN109472294A (en) * | 2018-10-15 | 2019-03-15 | 广州地理研究所 | A kind of recognition methods of urban water-body, device, storage medium and equipment |
| CN109740485A (en) * | 2018-12-27 | 2019-05-10 | 中国水利水电科学研究院 | Reservoir or dyke recognition methods based on spectrum analysis and depth convolutional neural networks |
| CN111695492A (en) * | 2020-06-10 | 2020-09-22 | 国网山东省电力公司电力科学研究院 | Method and system for detecting fishing hidden danger of power transmission line |
| CN114462542A (en) * | 2022-02-14 | 2022-05-10 | 长光禹辰信息技术与装备(青岛)有限公司 | Small target identification precision optimization method based on local difference analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542259A (en) * | 2011-12-29 | 2012-07-04 | 宁波工程学院 | Identification method for near-shore on-land water body |
| CN102708354A (en) * | 2011-12-31 | 2012-10-03 | 中国科学院遥感应用研究所 | Method for identifying golf course |
| CN103646246A (en) * | 2013-12-19 | 2014-03-19 | 交通运输部天津水运工程科学研究所 | Decision tree model based multispectral remote sensing image river information extraction method |
| CN105335761A (en) * | 2015-11-27 | 2016-02-17 | 郑州航空工业管理学院 | Remote-sensing image single-category information extraction method based on nearest neighbor method |
| CN105550695A (en) * | 2015-11-27 | 2016-05-04 | 郑州航空工业管理学院 | Object-oriented single-class classification method for remote sensing image |
-
2016
- 2016-11-02 CN CN201610944422.3A patent/CN106485239A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542259A (en) * | 2011-12-29 | 2012-07-04 | 宁波工程学院 | Identification method for near-shore on-land water body |
| CN102708354A (en) * | 2011-12-31 | 2012-10-03 | 中国科学院遥感应用研究所 | Method for identifying golf course |
| CN103646246A (en) * | 2013-12-19 | 2014-03-19 | 交通运输部天津水运工程科学研究所 | Decision tree model based multispectral remote sensing image river information extraction method |
| CN105335761A (en) * | 2015-11-27 | 2016-02-17 | 郑州航空工业管理学院 | Remote-sensing image single-category information extraction method based on nearest neighbor method |
| CN105550695A (en) * | 2015-11-27 | 2016-05-04 | 郑州航空工业管理学院 | Object-oriented single-class classification method for remote sensing image |
Non-Patent Citations (5)
| Title |
|---|
| WENKAI LI 等: "A Positive and Unlabeled Learning Algorithm for One-Class Classification of Remote-Sensing Data", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 * |
| 于晓升 等: "基于多特征融合的遥感图像河流目标检测算法", 《东北大学学报(自然科学版)》 * |
| 张绍明 等: "基于高分辨率遥感影像的内河航标自动检测方法", 《同济大学学报（自然科学版）》 * |
| 王凯峰 等: "基于单类SVM的遥感图像目标检测", 《计算机工程与应用》 * |
| 薄树奎 等: "面向对象的遥感影像单类分类", 《现代电子技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334909A (en) * | 2018-03-09 | 2018-07-27 | 南京天数信息科技有限公司 | Cervical carcinoma TCT digital slices data analysing methods based on ResNet |
| CN108334909B (en) * | 2018-03-09 | 2020-06-16 | 上海天数智芯半导体有限公司 | Cervical cancer TCT digital slice data analysis system based on ResNet |
| CN108920998A (en) * | 2018-04-14 | 2018-11-30 | 海南大学 | Single category based on Pasteur's coefficient pours down remote sensing image target area extracting method |
| CN108710862A (en) * | 2018-05-24 | 2018-10-26 | 河海大学 | A kind of high-resolution remote sensing image Clean water withdraw method |
| CN109472294A (en) * | 2018-10-15 | 2019-03-15 | 广州地理研究所 | A kind of recognition methods of urban water-body, device, storage medium and equipment |
| CN109740485A (en) * | 2018-12-27 | 2019-05-10 | 中国水利水电科学研究院 | Reservoir or dyke recognition methods based on spectrum analysis and depth convolutional neural networks |
| CN111695492A (en) * | 2020-06-10 | 2020-09-22 | 国网山东省电力公司电力科学研究院 | Method and system for detecting fishing hidden danger of power transmission line |
| CN114462542A (en) * | 2022-02-14 | 2022-05-10 | 长光禹辰信息技术与装备(青岛)有限公司 | Small target identification precision optimization method based on local difference analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106485239A (en) | One kind is using one-class support vector machines detection river mesh calibration method | |
| Huang et al. | A multidirectional and multiscale morphological index for automatic building extraction from multispectral GeoEye-1 imagery | |
| CN109871902B (en) | SAR small sample identification method based on super-resolution countermeasure generation cascade network | |
| CN102842032B (en) | Method for recognizing pornography images on mobile Internet based on multi-mode combinational strategy | |
| Huang et al. | Morphological building/shadow index for building extraction from high-resolution imagery over urban areas | |
| CN105184309B (en) | Classification of Polarimetric SAR Image based on CNN and SVM | |
| CN104036239B (en) | Fast high-resolution SAR (synthetic aperture radar) image ship detection method based on feature fusion and clustering | |
| CN104484681B (en) | Hyperspectral Remote Sensing Imagery Classification method based on spatial information and integrated study | |
| CN105335966B (en) | Multiscale morphology image division method based on local homogeney index | |
| CN103049763B (en) | Context-constraint-based target identification method | |
| CN104182763B (en) | A kind of floristics identifying system based on flower feature | |
| CN110135296A (en) | Airfield runway FOD detection method based on convolutional neural networks | |
| CN109325960A (en) | A kind of infrared cloud image cyclone analysis method and analysis system | |
| CN110309781B (en) | House damage remote sensing identification method based on multi-scale spectrum texture self-adaptive fusion | |
| CN108257151B (en) | PCANet image change detection method based on significance analysis | |
| CN109657610A (en) | A kind of land use change survey detection method of high-resolution multi-source Remote Sensing Images | |
| CN102622607A (en) | Remote sensing image classification method based on multi-feature fusion | |
| CN106295124A (en) | Utilize the method that multiple image detecting technique comprehensively analyzes gene polyadenylation signal figure likelihood probability amount | |
| CN108898065A (en) | Candidate regions quickly screen and the depth network Ship Target Detection method of dimension self-adaption | |
| CN109063754A (en) | A kind of remote sensing image multiple features combining classification method based on OpenStreetMap | |
| CN108664838A (en) | Based on the monitoring scene pedestrian detection method end to end for improving RPN depth networks | |
| CN106373146A (en) | Target tracking method based on fuzzy learning | |
| CN104252625A (en) | Sample adaptive multi-feature weighted remote sensing image method | |
| CN103729651A (en) | Hyperspectral remote sensing image classification method based on manifold neighbor measurement through local spectral angles | |
| Kalkan et al. | Comparison of support vector machine and object based classification methods for coastline detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 450000 Middle Road, 27 District University, Zhengzhou, Henan Province, No. 2 Applicant after: Zhengzhou Institute of Aeronautical Industry Management Address before: 450046 15 Wen Yuan Xi Road, Zheng Dong New District, Zhengzhou, Henan. Applicant before: Zhengzhou Institute of Aeronautical Industry Management |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170308 |