CN107192642A - A kind of Microphysical model construction techniques method of Atmospheric particulates - Google Patents
A kind of Microphysical model construction techniques method of Atmospheric particulates Download PDFInfo
- Publication number
- CN107192642A CN107192642A CN201710376801.1A CN201710376801A CN107192642A CN 107192642 A CN107192642 A CN 107192642A CN 201710376801 A CN201710376801 A CN 201710376801A CN 107192642 A CN107192642 A CN 107192642A
- Authority
- CN
- China
- Prior art keywords
- image
- microphysical
- aerosol
- model
- particulate
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume, or surface-area of porous materials
Abstract
The invention discloses a kind of Microphysical model construction techniques method of Atmospheric particulates.Electron microscope image of this method based on Atmospheric particulates, is split by the image of multiple features average drifting, determines exact position and region of the particulate matter on MIcrosope image first.Abundant morphological feature is obtained based on morphological properties section, accurate geological information is obtained using angle point and edge detecting technology, corresponding textural characteristics is generated by gray level co-occurrence matrixes, segmentation precision is lifted with reference to various features.According to microscope engineer's scale, obtain length and width, boundary length, inside/outside of the different components on image and connect the geological informations such as radius of circle.With reference to auxiliary informations such as chemical analysis, volume size, surface roughnesses, with reference to the morphological parameters of Typical particle thing, using image shadow recovery depth information, three-dimensional Microphysical model is directly reconstructed out from single-view two-dimensional projection image.By way of generating multiple Microphysical models and optimizing, to obtain more real analog result.
Description
Technical field
The present invention relates to a kind of Microphysical model construction techniques method of Atmospheric particulates, for weather such as gray haze and sand and dust
The microphysics chemical characteristic simulation of lower Atmospheric particulates.
Background technology
Particulate matter in Real Atmosphere has that form is random, composition is various, the features such as hybrid mode is complicated, its microphysics
Chemical characteristic is to restrict one of key factor of atmospheric parameter remote-sensing inversion and Radiative Forcing Evaluation accuracy.It is gentle in remote-sensing inversion
Wait in assessing, be volume equivalent sphere or multilayered sphere (Bohren, 2008) typically by the Microphysical model simplification of Atmospheric particulates.
But, based on the inversion result of this simplified model, have compared with actual observation data larger error (et
al.,2012)。
In recent years, increasing research focuses on how to simulate the airborne particulate for being closer to actual observation result
The Microphysical model of thing, so as to obtain more accurate optical characteristics (Adachi et al., 2010).Based on solid form
It is assumed that there are some researchs that Sand Dust Aerosol particle is assumed to be into spheroid, black carbon aerosols particle is assumed to be many pelletizings
Cluster carries out the simulation of microphysics chemical characteristic, obtains than the spherical optical diffusion characteristic assumed closer to actual measurement
(Nousiainen et al.,2006；Liu et al.,2008；Liu et al.,2013；Dong et al.,2015).But
It is that the optical characteristics result that the aerosol particle Microphysical modeling assumed based on this conventional solid form is obtained is still
There is certain difference with measured result, it is impossible to meet actual demand.
In order to simulate Atmospheric particulates characteristic closer to the truth, recent research begins to focus on no solid form
The Microphysical model construction of the Atmospheric particulates of constraint.Liu et al. (2015) are in black carbon aerosols particle Microphysical model structure
Telescopic effect (necking) is added during building, the microcosmic physicochemical property and the optical scattering that are more nearly actual measurement are obtained to simulate
Characteristic.Zubko et al., (2015) are constructed with random form and rough surface using irregular aggregate particles method
Particulate Microphysical model and optical scattering model, simulation calculate the optical characteristics for obtaining and being more nearly actual measurement.Wu et
Al. (2016) assume that each small particles in cluster also have non-spherical morphology, and simulation is obtained than many pelletizing clustering models more
Close to the black carbonaceous amount absorption cross-section of actual measurement.
At the same time, in the laboratory that some possess experiment condition, researcher starts to use transmission electron microscope(TEM) figure
As the method for three-dimensional reconstruction, to obtain the particulate form for being more nearly actual conditions.Lindqvist et al.
(2014) the transmission electron microscope(TEM) picture construction stereogram using various visual angles is proposed, the three-dimensional of particulate is directly obtained
Micromodel, to simulate the optical diffusion characteristic for being more nearly actual measurement.But, the technology is for sampling of aerosol and electronic transmission
The requirement of microscope work is higher, and the acquisition of particularly various visual angles electron microscope image data is more difficult, limits to larger.
Therefore, it is more nearly in continuous simulation in the research of actual measurement physics and chemistry optical characteristics, it is necessary to study a kind of direct base
In the aerosol particle thing scattering model construction method of actual measurement, the single two-dimensional projection that can be obtained by transmission electron microscope(TEM)
Image relatively accurately sets up the Microphysical model of aerosol particle thing, further improves the aerosol scattering of solid form constraint
Modeling method, the optical characteristics that can be preferably applied for amorphous aerosol particle thing is calculated and analyzed, and improves atmospheric parameter
The precision that remote-sensing inversion and Radiative Forcing are assessed.
The content of the invention
A kind of aerosol particle form from actual measurement is proposed, the spherical of routine, ellipsoid, cluster etc. is not preset fixed
Form is limited, the method for directly simulating amorphous aerosol particle microphysics chemical characteristic.Based on typical aerosol particle
Two-dimensional projection on transmission electron microscope(TEM) image, chemical analysis, volume size and the rough surface journey obtained with reference to actual measurement
The auxiliary parameters such as degree, it is considered to the physicochemical property of different type particulate, simulate the Microphysical of amorphous particulate
Model.
Comprise the following steps that：
(1) the aerosol particle micro image obtained based on laboratory facilities such as electron microscopes, is split first with image
Technology determines exact position and region of the particulate on MIcrosope image, extracts the geometry letter of different aerosol compositions
Breath.
(2) image classification of object-oriented is carried out using the means such as priori and artificial interpretation, cutting object phase is determined
Corresponding aerosol chemistry composition.
(3) image classification result is based on, with reference to auxiliary informations such as volume size, surface roughness and water content, figure is utilized
As shadow recovery depth information, then three-dimensional Microphysical model is directly reconstructed out from single-view two-dimensional projection image.
(4) by interactive meanses, the different chemical analysis of aerosol are introduced under the conditions of different wave length and relative humidity etc.
The physical and chemical parameters such as complex refractive index, so as to build the Microphysical model of particulate.
(5) by generating multiple Microphysical models, optimize and obtain more general aerosol particle Microphysical model,
Solve that the result and actual shape of aerosol three-dimensional reconstruction may be caused due to sampling of aerosol and microscopic angle
The problem of state is inconsistent, to obtain more real analog result.
(6) the Microphysical model of the particulate based on different time, research particulate size, microscopic pattern,
Physicochemical characteristics change rule in the characteristics such as chemical composition, hybrid mode and moisture absorption growth, analysis aerosol generation ageing process
Rule.
Brief description of the drawings
Fig. 1 for typical aerosol particle MIcrosope image (from left to right：Flue dust, sand and dust, sea salt)；
Fig. 2 a be based on morphological properties section obtain flue dust morphological feature (be from top to bottom area, the moment of inertia, standard
Difference；From left to right for 1,3,5,7,9)；
Fig. 2 b are the sand and dust morphological feature obtained based on morphological properties section；
Fig. 2 c are the sea salt morphological feature obtained based on morphological properties section；
Fig. 3 is the image segmentation result using adaptive multiple features average drifting；
Fig. 4 is the depth information that three-dimensional reconstruction obtains aerosol particle；
Fig. 5 is the general aerosol particle Microphysical model by multiple simulative optimization.
Embodiment
Below, refer to the attached drawing, is more fully illustrated to the present invention, shown in the drawings of the exemplary implementation of the present invention
Example.However, the present invention can be presented as a variety of multi-forms, the exemplary implementation for being confined to describe here is not construed as
Example.And these embodiments are to provide, so that the present invention is fully and completely, and it will fully convey the scope of the invention to this
The those of ordinary skill in field.
1st, aerosol microcosmic image is split
Due to particulate usually not solid form, geometry distribution is random, complicated various.Therefore, it is of the invention
The aerosol particle micro image obtained based on electron microscope, first by using the image segmentation side of multiple features average drifting
Method, to determine exact position and region of the particulate on MIcrosope image.Wherein, obtained based on morphological properties section
Abundant morphological feature, is obtained accurate geological information using angle point and edge detecting technology, is generated using gray level co-occurrence matrixes
Corresponding textural characteristics, contribute to the lifting of segmentation precision with reference to various features.
(1) expanding morphology attribute section (EAPs) is the form properties feature section using different type attributive character
(APs) it is superimposed expanded application.Each AP is the calculating that each characteristic layer (FRs) obtained after being converted based on image carries out feature.
EAP={ AP (FR1),AP(FR2),...,AP(FRc)} (1)
AP algorithms are a kind of feature extraction algorithms proposed on the basis of morphological properties filtering, and its basic thought is profit
Image is filtered with a series of morphological properties wave filter of different attributes to extract the structural information of image, by not
Integrated with attribute filter result, approximately can comprehensively describe the space geometry structure of image.For a width gray level image
F, according to certain rule T, AP algorithms carry out n form properties roughening computing and n feature refinement computing.
Wave filter carries out two-value prediction based on given reference value λ when performing AP algorithms to image attributes, if compared
Compared with property value be more than reference value λ, then image-region keep it is constant, conversely then set image intensity value be proximate region value,
So as to merge adjacent region.Therefore, a series of reference threshold { λ 1, λ 2 ... λ n } need to be set by calculating each AP, to carry out one
The attribute roughening and refinement computing of series.
In AP algorithms, the selection of attribute has diversity, including an attribute related to region shape, such as area, external
Square, the moment of inertia etc., also including the related attribute of the area grayscales such as gray average, entropy and standard deviation.It is different by selecting
Attribute, AP algorithms just can different types of feature in capture images.
Before isolated component and principal component component extraction EAPs features, integer processing is carried out to component data, by data
Value is stretched to the scope of [0,1000], with suitable for EAP algorithm requirements.According to the characters of ground object in experimental data region used, sheet
The form properties feature that selected works are selected includes area, standard deviation, the moment of inertia, and set property parameters λ value is respectively：λ a=
[200,400,600,1000], λ s=[20,30,40,50], λ i=[0.2,0.3,0.4,0.5].Test and tie for comparative analysis
Really, 4 wave bands original to image data carry out APs feature extractions, and each AP includes 27 characteristic layers, and 4 wave bands include 108 altogether
2 isolated component data after ICA is converted and PCA is converted are carried out EAPs feature extractions, 2 components are respectively wrapped by individual characteristic layer
Include 54 characteristic layers.
(2) the mean shift algorithms that segmentation is used, its essence is that feature space is gathered according to different standards
Class, if characteristic vector set Sd={ sk, k=1,2 ... }, wherein s=[ss, sr] T of the d dimensions of sampled data formation, general empty
Between domain vector ss be 2 dimensions, Range domains vector sr dimension is set to p, then d=p+2, in the set, probability density function
Parzen windows are estimated as
In above formula, bandwidth matrices H can be simplified expression by bandwidth factor h, H=h2I, at the same using profile function k come
Expression kernel function K (x)=k (| | x | |2), then formula (3) can be expressed as
By the separability of kernel function,
Wherein, C is normalization constant,WithThe different wideband coefficients in spatial domain and Range domains are represented respectively.
Density Estimator is meant that：It assign the karyomerite average value of a function centered on each data point as the data point
The estimate of probability density function, or：Kernel estimates are to calculate to be estimated a weighting in the window centered on being estimated a little
Local average.Bandwidth parameter h determines the size of the size of window, i.e. local neighborhood.
Improved mean shift algorithm considers spectral information (red, green, blue) and texture information in original basis.Basic
In the case that formula (5) is constant, mainly its kernel function is improved, formula after multiple features parameter is added as follows：
hs,hspe,ht- control smooth resolution
C-normaliztion constant
S-Spatial Dimension, two dimensional image
Spe-spectral signature (dimension), spe=1 represent that this is a gray-scale map, and spe=3 represents RGB color figure
T-textural characteristics (dimension), the textural characteristics of image are represented by gray level co-occurrence matrixes
2nd, the aerosol microcosmic image classification of object-oriented
Based on the training sample and priori being obtained ahead of time, the bulky grain aerosol such as sand and dust has with corner angle
Cluster form that there are the little particle aerosols such as irregular geometric shape, black carbon multiple bead particles mutually to assemble etc., knot
The means such as artificial interpretation are closed, the image classification of object-oriented is carried out, determines the corresponding aerosol chemistry composition of cutting object.
Carried out using regularization least square method (Regularized least-squares classification)
Object oriented classification, is obtained initial results, is then finely adjusted using interactive meanses.
For training sample { (x1, y1) ..., (xn, yn), regularization least square method is empty in reproducing kernel Hilbert
Between Hk(k is kernel function) so that function f meets following strategy：
Wherein, function f solution isAssuming that Ki,j=k (xi,xj), it is then that f* (x) both sides are flat
Side, can be obtainedCarrying it into minimum strategy then has：
By derivation, make derivative be equal to 0, solve equation:
(K+ λ nI) c=y (9)
Accordingly, for each point in training sample, grader is all set up using other n-1 point, classification is calculated and misses
Difference, constantly adjustment λ value are minimum until the error of grader.
3rd, the Microphysical model construction of aerosol particle
According to microscopical engineer's scale, obtain length and width of the different aerosol compositions on MIcrosope image, boundary length,
Inside/outside connects the geological informations such as radius of circle.With reference to auxiliary informations such as chemical analysis, volume size, surface roughnesses, with reference to typical case into
The morphological parameters of part particulate, using image shadow recovery depth information, are directly reconstructed from single-view two-dimensional projection image
Go out three-dimensional Microphysical model.
It is by aerosol when more because in MIcrosope image shooting process, direction of illumination and intensity are not fixed
Outwards project, directly obtained using the SFS methods (Shape from shaping) of traditional shadow recovery three-dimensional structure behind grain
Effect and bad.Therefore, for classification results above, background is adjusted to dead color by the adjustment of advanced column hisgram, and aerosol
Particle is then adjusted to light tone.The brightness for being additionally, since aerosol edge is generally larger, therefore by introducing in the middle of the solution of priori form
The situation of depression.
Priori form mainly includes：Sea salt is generally prism, typically based on quadrangular, six prisms etc.；Flue dust is generally spheroid,
But the particle size difference obtained may be larger simultaneously；Sand and dust are generally irregular form, and corner angle are clearly demarcated, to obtain completely true
Microphysical model it is extremely difficult, be generally fitted to the forms such as spheroid or prism, add portion forms change.
4th, general Microphysical model generation and application
The problem of due to sampling of aerosol and microscopic angle, the result and reality of aerosol three-dimensional reconstruction may be caused
Border state is inconsistent, it is therefore desirable to be modified based on priori.Also, optimized by generating multiple Microphysical models
Mode, to obtain more real analog result.
Due in same time and region there is same type of aerosol particle to be morphologically consistent substantially, therefore
Based on multiple Microphysical models, the uncertainty of single aerosol particle can be corrected, more general aerosol particle is obtained
Microphysical model, the simulation for optical diffusion characteristic.
Multiple particles are amplified to a size, three-dimensional overlay average treatment is then carried out, multiple forms are obtained in three-dimensional
Probability of occurrence in space, by setting threshold value, will be greater than being equal to the spatial point of threshold value as the part of universal model, and be less than
Threshold value does not then consider.By the optimization of multiple models, aerosol particle people's universal model is obtained.
One of method innovation point proposed by the present invention is, based on transmission electron microscope(TEM) image, to utilize image Segmentation Technology
The physicochemical characteristicses of particulate are extracted, with reference to corresponding auxiliary parameter and priori, based on individual figure of image shadow recovery
The depth information of picture, the Microphysical model of particulate is constructed using three-dimensional reconstruction.
The present invention be directed to the aerosol particle Microphysical model building method based on MIcrosope image, with following nature
The technical characteristic of rule, this method solve the technical problem of aerosol particle Microphysical model construction, and be not belonging to Patent Law
The rules and methods of intellection described in 2 sections of 25.
Claims (4)
1. claim 1：A kind of aerosol particle form from actual measurement is proposed, the spherical of routine, ellipsoid, group is not preset
The solid forms such as cluster are limited, the method for directly simulating amorphous aerosol particle microphysics chemical characteristic.Specific steps are such as
Under：(1) two-dimensional projection based on typical aerosol particle on transmission electron microscope(TEM) image, is determined using image Segmentation Technology
Exact position and region of the particulate on MIcrosope image, extract the geological information of different aerosol compositions.(2) utilize
The means such as priori and artificial interpretation carry out the image classification of object-oriented, determine the corresponding aerosol chemistry of cutting object
Composition.(3) auxiliary informations such as volume size, surface roughness and water content are combined, using image shadow recovery depth information, so
Afterwards three-dimensional Microphysical model is directly reconstructed out from single-view two-dimensional projection image.(4) by interactive meanses, aerosol is introduced different
The physical and chemical parameters such as complex refractive index of the chemical analysis under the conditions of different wave length and relative humidity etc., so as to build particulate
Microphysical model.(5) more general aerosol particle Microphysical model is obtained by generating multiple Microphysical model optimizations
Analog result.(6) the Microphysical model of the particulate based on different time, research particulate size, microscopic pattern,
Physicochemical characteristics change rule in the characteristics such as chemical composition, hybrid mode and moisture absorption growth, analysis aerosol generation ageing process
Rule.
2. claim 2：Step (1) as claimed in claim 1, based on typical aerosol particle in transmission electron microscope(TEM) figure
As upper two-dimensional projection, exact position and region of the particulate on MIcrosope image are determined using image Segmentation Technology,
Extract the geological information of different aerosol compositions.Towards the MIcrosope image of particulate, floated by using multiple features average
The image partition method of shifting, to determine exact position and region of the particulate on MIcrosope image.Wherein, using a variety of
The method that feature is combined is used to lift segmentation precision, main to include obtaining abundant morphological feature based on morphological properties section,
Accurate geological information is obtained using angle point and edge detecting technology, corresponding textural characteristics are generated using gray level co-occurrence matrixes.
Then, described step (2) is required by right 1, the image of object-oriented is carried out using the means such as priori and artificial interpretation
Classification, determines the corresponding aerosol chemistry composition of cutting object.
3. claim 3：Step (3) as claimed in claim 1, it is auxiliary with reference to volume size, surface roughness and water content etc.
Supplementary information, using image shadow recovery depth information, then directly reconstructs out three-dimensional Microphysical from single-view two-dimensional projection image
Model.According to microscopical engineer's scale, obtain length and width of the different aerosol compositions on MIcrosope image, boundary length, it is interior/
The geological informations such as circumradius.With reference to auxiliary informations such as chemical analysis, volume size, surface roughnesses, with reference to typical composition
The knowledge such as the morphological parameters and prior model of particulate, using image shadow recovery depth information, throw from single-view two dimension
Shadow image directly reconstructs out three-dimensional Microphysical model.Step (4) as claimed in claim 1, by interactive meanses, introduces gas molten
The physical and chemical parameters such as complex refractive index of the glue difference chemical analysis under the conditions of different wave length and relative humidity etc., so that it is molten to build gas
The Microphysical model of micelle.
4. claim 4：Step (5) as claimed in claim 1 and (6), are obtained by generating multiple Microphysical model optimizations
More general aerosol particle Microphysical model simulation results.Multiple particles are amplified to a size, then carried out three-dimensional
Superposed average processing, obtains probability of occurrence of multiple forms in three dimensions, by setting threshold value, will be greater than equal to threshold value
Spatial point and then not considering as the part of universal model less than threshold value.By the optimization of multiple models, aerosol is obtained
Grain people's universal model.Then, the Microphysical model of the particulate based on different time, it is research particulate size, micro-
See physicochemical characteristics in the characteristics such as form, chemical composition, hybrid mode and moisture absorption growth, analysis aerosol generation ageing process
Changing rule.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710376801.1A CN107192642B (en) | 2017-05-25 | 2017-05-25 | A kind of Microphysical model building method of atmospheric aerosol particle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710376801.1A CN107192642B (en) | 2017-05-25 | 2017-05-25 | A kind of Microphysical model building method of atmospheric aerosol particle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107192642A true CN107192642A (en) | 2017-09-22 |
| CN107192642B CN107192642B (en) | 2019-11-08 |
Family
ID=59874817
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710376801.1A Active CN107192642B (en) | 2017-05-25 | 2017-05-25 | A kind of Microphysical model building method of atmospheric aerosol particle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107192642B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614851A (en) * | 2018-10-26 | 2019-04-12 | 上海市环境监测中心(上海长三角区域空气质量预测预报中心) | The image detecting method of atmosphere pollution particulate matter morphological parameters |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706966A (en) * | 2009-11-06 | 2010-05-12 | 上海第二工业大学 | Method for three-dimensional reconstruction of porous medium on basis of two-dimensional images and multi-point statistical method |
| CN102155909A (en) * | 2010-12-17 | 2011-08-17 | 华中科技大学 | Nano-scale three-dimensional shape measurement method based on scanning electron microscope |
| CN102254354A (en) * | 2011-06-27 | 2011-11-23 | 中国科学院武汉岩土力学研究所 | Method for implementing microstructure surface morphology three-dimensional visualization of soil |
| CN102254303A (en) * | 2011-06-13 | 2011-11-23 | 河海大学 | Methods for segmenting and searching remote sensing image |
| CN106383207A (en) * | 2016-10-26 | 2017-02-08 | 中国科学院合肥物质科学研究院 | Computing method applied to atmospheric aerosol mass concentration horizontal route distribution |
-
2017
- 2017-05-25 CN CN201710376801.1A patent/CN107192642B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101706966A (en) * | 2009-11-06 | 2010-05-12 | 上海第二工业大学 | Method for three-dimensional reconstruction of porous medium on basis of two-dimensional images and multi-point statistical method |
| CN102155909A (en) * | 2010-12-17 | 2011-08-17 | 华中科技大学 | Nano-scale three-dimensional shape measurement method based on scanning electron microscope |
| CN102254303A (en) * | 2011-06-13 | 2011-11-23 | 河海大学 | Methods for segmenting and searching remote sensing image |
| CN102254354A (en) * | 2011-06-27 | 2011-11-23 | 中国科学院武汉岩土力学研究所 | Method for implementing microstructure surface morphology three-dimensional visualization of soil |
| CN106383207A (en) * | 2016-10-26 | 2017-02-08 | 中国科学院合肥物质科学研究院 | Computing method applied to atmospheric aerosol mass concentration horizontal route distribution |
Non-Patent Citations (3)
| Title |
|---|
| H.LINDQVIST ET AL: "Single scattering by realistic,", 《ATOMS.CHEM.PHYS.DISCUSS.》 * |
| HANNAKAISA LINDQVIST ET AL: "Optical modeling of vesicular volcanic ash particles", 《JOURNAL OF QUANTITATIVE SPECTROSCOPY & RADIATIVE TRANSFER》 * |
| 孙小炜 等: "Mean_Shift图像分割的快速算法", 《测控技术》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109614851A (en) * | 2018-10-26 | 2019-04-12 | 上海市环境监测中心(上海长三角区域空气质量预测预报中心) | The image detecting method of atmosphere pollution particulate matter morphological parameters |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107192642B (en) | 2019-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109147254B (en) | Video field fire smoke real-time detection method based on convolutional neural network | |
| CN104484668B (en) | A kind of contour of building line drawing method of the how overlapping remote sensing image of unmanned plane | |
| CN109271856A (en) | Remote sensing image object detection method based on expansion residual error convolution | |
| CN109508710A (en) | Based on the unmanned vehicle night-environment cognitive method for improving YOLOv3 network | |
| CN110032949A (en) | A kind of target detection and localization method based on lightweight convolutional neural networks | |
| CN105335966B (en) | Multiscale morphology image division method based on local homogeney index | |
| CN108596108B (en) | Aerial remote sensing image change detection method based on triple semantic relation learning | |
| CN109255831A (en) | The method that single-view face three-dimensional reconstruction and texture based on multi-task learning generate | |
| CN109816024A (en) | A kind of real-time automobile logo detection method based on multi-scale feature fusion and DCNN | |
| CN109284704A (en) | Complex background SAR vehicle target detection method based on CNN | |
| CN113486764B (en) | Pothole detection method based on improved YOLOv3 | |
| CN113160062B (en) | Infrared image target detection method, device, equipment and storage medium | |
| CN101901343A (en) | Remote sensing image road extracting method based on stereo constraint | |
| CN106407951B (en) | A kind of night front vehicles detection method based on monocular vision | |
| CN108399625A (en) | A kind of SAR image orientation generation method generating confrontation network based on depth convolution | |
| CN110309781A (en) | Damage remote sensing recognition method in house based on the fusion of multi-scale spectrum texture self-adaption | |
| CN111126278A (en) | Target detection model optimization and acceleration method for few-category scene | |
| CN110222586A (en) | A kind of calculating of depth of building and the method for building up of urban morphology parameter database | |
| CN109902629A (en) | A kind of real-time vehicle target detection model under vehicles in complex traffic scene | |
| CN112348758B (en) | Optical remote sensing image data enhancement method and target identification method | |
| CN108921817A (en) | A kind of data enhancement methods for skin disease image | |
| CN110909623A (en) | Three-dimensional target detection method and three-dimensional target detector | |
| CN115512247A (en) | Regional building damage grade assessment method based on image multi-parameter extraction | |
| CN106600613A (en) | Embedded GPU-based improved LBP infrared target detection method | |
| CN115272876A (en) | Remote sensing image ship target detection method based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |