CN112632661A - SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm - Google Patents

SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm Download PDF

Info

Publication number
CN112632661A
CN112632661A CN202011467882.4A CN202011467882A CN112632661A CN 112632661 A CN112632661 A CN 112632661A CN 202011467882 A CN202011467882 A CN 202011467882A CN 112632661 A CN112632661 A CN 112632661A
Authority
CN
China
Prior art keywords
picture
pictures
point
modified asphalt
points
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
Application number
CN202011467882.4A
Other languages
Chinese (zh)
Inventor
胡魁
俞才华
张桃利
栗启
陈玉静
聂思雨
王悦
陈桂香
王迪
贾艺玮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Henan University of Technology
Original Assignee
Henan University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Henan University of Technology filed Critical Henan University of Technology
Priority to CN202011467882.4A priority Critical patent/CN112632661A/en
Publication of CN112632661A publication Critical patent/CN112632661A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Geometry (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Structural Engineering (AREA)
  • Computational Mathematics (AREA)
  • Civil Engineering (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Architecture (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a method based on intelligent recognition algorithmSBSThe modified asphalt three-dimensional microstructure reconstruction method comprises the steps of preparing asphalt slices by freezing,SBSshooting the modified asphalt fluorescent picture, selecting nine pictures of each layer of slices, and sequentially splicing the nine pictures according to the shooting sequence to enable the nine pictures to be spliced into a complete picture; the splicing process comprises the steps of establishing a differential pyramid for a picture, determining feature points in the picture, and establishing descriptors of the feature points after setting a main direction for the feature points; splicing two adjacent pictures by matching descriptors of the same characteristic point, so that nine pictures of each layer of slices are spliced into a panoramic picture; performing three-dimensional modeling and materialization processing on panoramic pictures of all slices to obtain panoramic picturesSBSAnd (3) modifying the asphalt three-dimensional model. The method is as followsNow showSBSAnd the reconstruction of the three-dimensional microstructure of the modified asphalt provides reference for the optimization design of the asphalt material.

Description

SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm
Technical Field
The invention relates to the field of road asphalt building materials, in particular to a method for reconstructing a three-dimensional microstructure of SBS modified asphalt based on an intelligent recognition algorithm.
Background
Since ancient times, technological innovation has been pushing forward the development of human society with an irreversible and irresistible force. Currently, a new technological revolution is being developed vigorously, profoundly affecting and changing various industries, and the traditional road traffic is no exception. For a long time, the structural design of the asphalt pavement is based on the theory of an elastic layered system and belongs to the category of continuous media; the design of the asphalt pavement material focuses on the overall performance of the composition material, and the macroscopic performance evaluation is taken as a main index; the method is an ideal engineering model under the condition of limited calculation and analysis means and material recognition visual field. However, with the deep understanding of the complexity of the road surface, the original continuous medium system is particularly limited when explaining the damage and damage mechanism of the asphalt pavement material and realizing the synergistic design of the structure and the material. In fact, if calculated according to the mass ratio, the granular aggregate accounts for more than 90% and the volume ratio accounts for more than 85% in the composition of the asphalt pavement material, the rutting diseases of the asphalt pavement are mainly formed by the flowing deformation of the aggregate particles, the pavement crack diseases are mainly developed by micro-damage among the aggregate particles, and the pavement water damage diseases are mainly caused by the peeling of asphalt on the surfaces of the aggregate particles, so the asphalt pavement material is a typical particulate matter system, and the service performance of the asphalt pavement material is closely related to the shape, characteristics, grading, interparticle interfaces and other microscopic parameters of the particulate material.
The traditional asphalt pavement material research focuses on the correlation of mechanical properties of macro-micro road of asphalt and aggregate in the continuous medium category, while the unstable phenomenon of multiphase reaction in the processing process and the physical and chemical problems of complex microstructure evolution in the service process of the asphalt material are ignored. Practical experience has shown that the development of rutting, cracking, water damage in asphalt pavements is closely related to the microscopic properties of the asphalt material and the material response. Numerous studies have shown that the microstructure of modified bitumen is closely related to the macroscopic properties, which are also considered to be an outward manifestation of the internal structure. The fluorescence technology is characterized based on the difference of the fluorescence response of the modifier and the matrix asphalt, and becomes a test means for researching the effectiveness and the great potential of the microstructure. Non-linear polymers have been favored in recent research because they have significant advantages over linear molecules in terms of solubility, viscosity, stability, and versatility. The method for researching complex nonlinear polymerization reaction and molecular structure and performance by adopting a computer simulation method not only has important scientific value, but also has important guiding significance for controlling reaction and accelerating application of materials in experiments.
As the SBS association and the matrix asphalt show the difference of bright yellow and dark under a fluorescence microscope, the fluorescence microscopic analysis means is the most direct and valuable method for researching the microstructure of the SBS modified asphalt. In previous researches, a fluorescence micrograph provides a method for testing the average value of the swelling diameter of the SBS modifier in the modified asphalt, a method for optimizing the compatibility of the SBS modifier and the matrix asphalt and the optimal mixing amount of the modifier, and a meaningful quantitative evaluation result of the microstructure is obtained. However, if the physical and chemical problems of the microstructure evolution of the asphalt material during the processing, the nonlinear rheology of the asphalt material and the mechanism of the instability phenomenon of asphalt molecule processing are further studied, the conventional planar picture-based evaluation method is difficult to realize.
Disclosure of Invention
The invention aims to provide a three-dimensional microstructure reconstruction method of SBS modified asphalt based on an intelligent recognition algorithm, which is used for overcoming the problem that the microstructure of SBS modified asphalt cannot be effectively evaluated by the traditional method based on a plane picture.
In order to realize the task, the invention adopts the following technical scheme:
an SBS modified asphalt three-dimensional microstructure reconstruction method based on an intelligent recognition algorithm comprises the following steps:
freezing to prepare asphalt slices, comprising:
preparing SBS modified asphalt, preheating to a flowing state, pouring into a prefabricated test mold, inserting a steel needle at any three corners of the test mold for positioning, and freezing an SBS modified asphalt test piece at the temperature of minus 10 ℃ to minus 5 ℃ to enable the modified asphalt test piece to be in a solid state;
SBS modified asphalt fluorescence picture shooting, including:
pulling out the positioning steel needle, taking out the SBS modified asphalt test piece, cutting the solid modified asphalt test piece into N +2 layers of slices with the thickness of 1mm by using an ultrasonic rubber cutting knife, removing the two layers of slices on the upper surface and the lower surface, and numbering the rest N slices from bottom to top according to the number of 1-N; sequentially placing all the slices on a slide glass of a fluorescence microscope objective table for positioning, continuously shooting nine pictures at adjacent positions and numbering the pictures for the slices of each layer from a selected initial position according to a nine-square grid arrangement mode under the condition of eliminating bubbles between the slide glass and asphalt, and ensuring that the two adjacent pictures have partial overlapping areas;
picture stitching, including:
selecting nine pictures of each layer of slices, and sequentially splicing the nine pictures according to the photographing sequence to splice the nine pictures into a complete picture; the splicing process comprises the steps of establishing a differential pyramid for a picture, determining feature points in the picture, setting a main direction for the feature points, and then constructing descriptors of the feature points; splicing two adjacent pictures by matching descriptors of the same characteristic point, so that nine pictures of each layer of slices are spliced into a panoramic picture;
and respectively converting the panoramic pictures of all the slices into gray level images and binary images, establishing a three-dimensional model, performing triangular patch quantity reduction on the established three-dimensional model, dividing a three-dimensional model surface grid again, materializing the triangular patch model in the three-dimensional model, diagnosing the defects and repairing the defects to obtain the final materialized SBS modified asphalt three-dimensional model.
Further, the establishing a differential pyramid for the picture includes:
establishing a characteristic pyramid for the sliced pictures, wherein each rectangle in the characteristic pyramid represents one picture, the pictures with the same size form a group, one picture in the group is one layer, and different layers in the same group of pictures are formed by convolution of convolution kernels with different scales in the first layer in each group of pictures; any group of pictures in the characteristic pyramid are obtained by downsampling a previous group of pictures;
the formula for calculating the group number O and the layer number S of the characteristic pyramid is as follows:
O＝[log2(min(M,N))]-3
S＝n+3
in the above formula, M, N represents the length and width of the original picture, respectively, [ ] represents rounding, and n represents the number of the feature point pictures to be extracted;
the differential pyramid is the difference between two adjacent layers of pictures in the same group of the characteristic pyramid, the number of layers of each group of pictures in the differential pyramid is one layer less than that of the corresponding group of pictures in the characteristic pyramid, and the first layer of convolution kernel scale sigma in the differential pyramid0And the value of the parameter k is calculated by adopting the following formula:
k＝21/n
σ0＝1.52
in the first group, the first layer uses σ0As convolution scale, k σ is used for the second layer0As a convolution scale, the nth layer is represented by kn-1σ0As a convolution scale; and a second group of first layers with knσ0As a convolution scale, k for the second layern+1σ0As a convolution scale, the convolution scales of other layers of other groups, and so on.
Further, the determining the feature points in the picture includes:
calculating partial derivatives in three directions of a differential pyramid, including a picture length direction x, a picture width direction y and a picture kernel convolution direction sigma, wherein a point with a partial derivative of 0 is an extreme point;
the obtained extreme points are possibly picture feature points, and in order to further determine the positions of the feature points, the extreme points need to be accurately positioned, which includes:
in the differential pyramid, the pixel where the extreme point is located should be larger or smaller than the rest of the pixel points in the three directions, and before accurate positioning, the picture should be thresholded; reading the pixel value of each pixel point of the picture, and considering that the point cannot be an extreme point when the following formula is not met:
abs(val)≥0.5×D/n
in the above formula, abs (val) is the absolute value of the pixel point, D is an empirical parameter, 0.04 is taken, and n is the number of the feature point images to be extracted;
after thresholding of the picture, partial derivatives in three directions (X, y, sigma) are obtained for the whole view angle space, a pixel point with a partial derivative of 0 is obtained, and coordinates are set as X0(x0,y0,σ0)T(ii) a Precisely locating the extreme point of the picture by taking the picture pixel value function f (X) at the extreme point X0(x0,y0,σ0)TThe part is unfolded to obtain;
for all the extreme points found, if
Figure BDA0002835090440000041
The contrast of the point pair is considered to be low and not to be a feature point, wherein f (x) is a function of the pixel value of the picture; and eliminating the extreme points with edge effect for the rest extreme points, and taking the rest extreme points as feature points.
Further, the extreme point of the picture is accurately positioned, and the extreme point X is further accurately positioned0(x0,y0,σ0)TIs obtained by unfolding and comprises the following steps:
setting the extreme point at the viewing angleThe coordinate of space is X0(x0,y0,σ0)TThe picture pixel value function f (x) is expanded at this point:
Figure BDA0002835090440000042
in the above formula, f is a picture pixel function having three components of x, y, and σ, the superscript T denotes a transpose matrix,
Figure BDA0002835090440000043
calculating a partial derivative; converting the above equation into a vector representation:
Figure BDA0002835090440000051
wherein X ═ X, y, σ ]; the partial derivatives of f (x) are calculated to obtain:
Figure BDA0002835090440000052
if the partial derivative is 0, solving the above equation yields:
Figure BDA0002835090440000053
Figure BDA0002835090440000054
the extreme point is taken into f (X) to obtain the extreme value
Figure BDA0002835090440000055
Figure BDA0002835090440000056
Further, the eliminating extreme points with edge effects includes:
the curvature of the pixel value function f (x) is represented by a matrix H, which is expressed as follows:
Figure BDA0002835090440000057
wherein D isxx(x,y)、Dxy(x,y)、Dyx(x,y)、Dyy(x, y) are each
Figure BDA0002835090440000058
Figure BDA0002835090440000059
A matrix formed by the derivation results;
the eigenvalues are replaced by traces tr (H) and determinant det (H) of the matrix H; the H matrix has two eigenvalues, set as α and β, where α > β, and α ═ γ β, γ being a coefficient; the relationship of the characteristic values to Tr (H) and Det (H) is represented by the following formula:
Tr(H)＝Dxx(x,y)+Dyy(x,y)＝α+β
Det(H)＝Dxx(x,y)Dyy(x,y)-(Dxy(x,y))2＝αβ
if Det (H) < 0, i.e., α and β are off, then there is an edge effect at this point and it should be dropped; if Det (H) > 0, α and β have the same sign, and α > β, so γ > 1, the following formula is calculated:
Figure BDA0002835090440000061
if it is
Figure BDA0002835090440000062
Then the point is considered to have an edge effect and should be discarded; if it is
Figure BDA0002835090440000063
The point is considered to be the desired extreme point, i.e., the feature point.
Further, the setting of the main direction for the feature point includes:
taking the feature point as the center of a circle, taking 1.5 times of the convolution scale sigma of the difference pyramid picture where the feature point is located or is closest to the feature point as the radius, counting the gradient direction and the gradient amplitude of all pixel points in the circle, performing 1.5 sigma Gaussian filtering, weighting the pixel points at different positions in the circle, and taking the direction with the maximum amplitude as the main direction of the feature point.
Further, the building of the descriptors of the feature points and the splicing of two adjacent pictures by matching the descriptors of the same feature points includes:
selecting a square pixel area by taking the characteristic point as a center, dividing the area into 16 subregions with the same size, counting gradients of pixel points in the subregions in 8 directions in each subregion, and weighting the gradients by Gaussian filtering; each subregion has 8 directional gradients and 16 subregions, and the descriptor of each feature point is a 128-dimensional vector;
by matching the descriptors of the same feature points, the splicing of two adjacent pictures can be realized.
Compared with the prior art, the invention has the following technical characteristics:
the method can realize the digital reconstruction of the three-dimensional microstructure of the SBS modified asphalt under the fluorescent picture, can comprehensively evaluate the microstructure of the SBS modified asphalt, provides a basis for numerical calculation and finite element simulation, and has important engineering significance and application prospect.
Drawings
Fig. 1 (a) and (b) are a feature pyramid and a differential pyramid;
FIG. 2 is a convolution kernel convolution scale sigma value of different positions of a characteristic pyramid;
FIG. 3 is an assignment of feature point directions;
FIG. 4 is a technical scheme of the present invention;
FIG. 5 is a tomographic fluorescent scanning step of an asphalt section;
FIG. 6 is a nine-square grid layout and numbering of a SBS modified asphalt layer slice picture;
FIG. 7 shows the image before and after splicing in the Sudoku arrangement mode;
fig. 8 is a reconstruction of a three-dimensional microstructure.
Detailed Description
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings. It should be noted that the present invention is not limited to the following specific examples.
Referring to the attached drawings, the invention discloses a SBS modified asphalt three-dimensional microstructure reconstruction method based on an intelligent recognition algorithm, which comprises the following steps:
step 1, preparing asphalt slices by quick freezing.
The SBS modified asphalt is preheated to a flowing state, and poured into a prefabricated test mold, wherein the test mold adopted in the embodiment is a rectangular body with four side surfaces and the size of 15mm multiplied by 22mm, and the test mold is made of glass materials. And respectively inserting a steel needle at any three corners of the test mold for positioning, and freezing the SBS modified asphalt sample at the temperature of minus 10 ℃ to minus 5 ℃ to enable the modified asphalt sample to be solid.
And 2, shooting the SBS modified asphalt fluorescent picture.
Pulling out the positioning steel needle, taking out the SBS modified asphalt test piece, cutting the solid modified asphalt test piece into N +2 layers of slices with the thickness of 1mm by using an ultrasonic rubber cutting knife, removing the two layers of slices on the upper surface and the lower surface, and numbering the rest slices from bottom to top according to the number of 1-N; in this embodiment, the value of N is 20.
Sequentially placing all the slices on a slide glass of a fluorescence microscope objective table for positioning, continuously shooting nine pictures at adjacent positions and numbering the pictures for each layer of the slices according to a nine-square grid arrangement mode from a selected initial position under the condition of eliminating bubbles between the slide glass and asphalt, and ensuring that the two adjacent pictures have partial overlapping areas; the picture taken in this embodiment is about 70 μm to 90 μm outward from the center of the slice.
And step 3, splicing the pictures.
And selecting nine sliced pictures of each layer, and sequentially splicing the nine pictures according to the photographing sequence to splice the nine pictures into a complete picture. The image splicing comprises 4 steps of establishing a differential pyramid, accurately positioning feature points, determining the main direction of the feature points and constructing and matching feature point descriptors, and the specific process is as follows:
and 3.1, constructing a differential pyramid.
The key point of image splicing is detection and matching of characteristic points, and to find the characteristic points of the image, a real visual angle space must be established in a computer at first, and the real visual angle space is established in the computer, and the key point is that the characteristics of the real world scene, such as large and small near parts, clear and fuzzy far parts, are simulated. Thus, building a differential pyramid in a computer can simulate a real view angle space. For constructing the differential pyramid, the following scheme is adopted:
as shown in fig. 1, (a) of fig. 1 is a characteristic pyramid. Each rectangle of the feature pyramid represents one picture, the pictures with the same size form a group, and one picture in the group is a layer. Any group of pictures in the feature pyramid are obtained by downsampling a previous group of pictures, and downsampling means that a pixel point is taken at every other pixel point, which means that the side length of any group of pictures is half of that of the previous group of pictures. The effect is that the larger the group number, the smaller the picture size, which is a characteristic of the real viewing angle. The different layers in the same group of pictures are formed by convolving the first layer in each group of pictures by convolution kernels with different scales (variances), and the effect is that the definition degrees of the same group of pictures are different, so that the characteristics of near definition and far blur of the real world can be simulated. The formula for calculating the group number O and the layer number S of the characteristic pyramid is as follows:
O＝[log2(min(M,N))]-3
S＝n+3
in the above formula, M, N represents the length and width of the original picture, respectively, [ ] represents rounding, and n represents the number of feature point pictures to be extracted.
The difference pyramid is the difference between two adjacent layers of pictures in the same group of the feature pyramid. Therefore, the number of each group of picture layers of the differential pyramid is compared with that of the feature goldThe number of picture layers in the corresponding group of the character tower is one layer less, as shown in fig. 1 (b). In addition, the convolution kernel scale σ (variance) takes on the value shown in FIG. 2, where the first layer scale σ is0And the value of the parameter k is calculated by adopting the following formula:
k＝21/n
Figure BDA0002835090440000081
in the first group, the first layer is represented by σ, as shown in FIG. 20As convolution scale, k σ is used for the second layer0The n-th (the number of feature point images to be extracted) layer is kn-1σ0As a convolution scale. And a second group of first layers with knσ0As a convolution scale, k for the second layern+1σ0As a convolution scale. The convolution scale σ of the pictures at other positions in the feature pyramid can be analogized.
And 3.2, accurately positioning the characteristic points.
After the differential pyramid is established, a real view angle space in a computer is established, and at this time, the feature point position in the picture can be searched, the method is to solve the partial derivative in three directions (the picture length direction x, the picture width direction y and the picture kernel convolution direction sigma) of the differential pyramid, and the point with the partial derivative being 0 is an extreme point. The obtained extreme points may be the picture feature points to be found in the present scheme, and in order to further determine the positions of the feature points, the extreme points need to be accurately located, and the specific method is as follows.
In the differential pyramid, the extreme point should be located in a pixel larger or smaller than the remaining 26 pixels in the three directions, and the picture should be thresholded before accurate positioning.
Reading the pixel value of each pixel point of the picture, and considering that the point cannot be an extreme point when the following formula is not met:
abs(val)≥0.5×D/n
in the above formula, abs (val) is the absolute value of the pixel, D is 0.04, and n is the number of feature point images to be extracted. D is an empirical parameter and is set to 0.04 because when it is less than this value, it is easily affected by noise, thereby undermining the choice of the extreme point.
After thresholding of the picture, partial derivatives in three directions (X, y, sigma) are calculated for the whole view angle space to obtain pixel points with partial derivatives of 0, and coordinates are set as X0(x0,y0,σ0)T. Since the view space created by the computer is discrete (consisting of individual pixels), which means that the direct derivation may not be a true extreme, the picture extreme should be precisely located by locating the extreme X0(x0,y0,σ0)TThe process of unfolding the stent comprises the following steps:
firstly, the coordinate of the extreme point in the visual angle space is assumed to be X0(x0,y0,σ0)TThe pixel value function f (x) of the picture is expanded at this point:
Figure BDA0002835090440000091
in the above formula, f is a picture pixel function having three components of x, y, and σ, the superscript T denotes a transpose matrix,
Figure BDA0002835090440000101
calculating a partial derivative; converting the above equation into a vector representation:
Figure BDA0002835090440000102
wherein X ═ X, y, σ ]; the partial derivatives of f (x) are calculated to obtain:
Figure BDA0002835090440000103
if the partial derivative is 0, solving the above equation yields:
Figure BDA0002835090440000104
Figure BDA0002835090440000105
the extreme point is taken into f (X) to obtain the extreme value
Figure BDA0002835090440000106
Figure BDA0002835090440000107
After the above operation, a series of extreme points are obtained, and a part of the points with lower contrast should be cut off if the above operation is performed
Figure BDA0002835090440000108
The point pair is considered to be low in contrast and not a characteristic point.
After obtaining the extreme point where the contrast is high, the edge effect should also be removed:
the curvature of the pixel value function f (x) is represented by a matrix H, which is expressed as follows:
Figure BDA0002835090440000109
wherein D isxx(x,y)、Dxy(x,y)、Dyx(x,y)、Dyy(x, y) are each
Figure BDA0002835090440000111
Figure BDA0002835090440000112
And (5) obtaining a matrix formed by the derivative result.
The matrix may describe the curvature of the pixel, and in order to eliminate edge effects it is desirable that the curvature of the pixel curve in both the x and y directions differ not much, otherwise the point is likely to be an edge. The eigenvalues of the matrix are proportional to the curvature, and for the sake of simplicity of calculation, traces tr (H) and determinant det (H) of matrix H may be used instead of eigenvalues; the H matrix is a 2 × 2 matrix, and thus has two eigenvalues, set to α and β, where α > β, and α ═ γ β, γ being a coefficient; the relationship of the characteristic values to Tr (H) and Det (H) is represented by the following formula:
Tr(H)＝Dxx(x,y)+Dyy(x,y)＝α+β
Det(H)＝Dxx(x,y)Dyy(x,y)-(Dxy(x,y))2＝αβ
if Det (H) < 0, i.e., α and β are off-signs, then there is an edge effect at this point and it should be dropped. If Det (H) > 0, α and β have the same sign, and α > β, so γ > 1, the following formula is calculated:
Figure BDA0002835090440000113
if it is
Figure BDA0002835090440000114
Then the point is considered to have an edge effect and should be discarded; if it is
Figure BDA0002835090440000115
The point is considered to be the desired extreme point, i.e., the feature point. In addition, γ is usually 10, that is,
Figure BDA0002835090440000116
and 3.3, determining the main direction of the characteristic point.
After the feature points of the picture are screened out, the directions of the feature points should be given, and this step can be realized by counting the gradient magnitude of the feature points.
As shown in fig. 3, with the feature point as the center of a circle and 1.5 times of the convolution scale σ of the difference pyramid picture where the feature point is located or closest as the radius, the gradient direction and the gradient amplitude of all the pixel points in the circle are counted, and gaussian filtering of 1.5 σ is performed to weight the pixel points at different positions in the circle, and the direction with the largest amplitude is taken as the main direction of the feature point.
And 3.4, constructing the feature point descriptor.
Through the steps, the feature points are selected, the main directions of the feature points are counted, and then descriptors of the feature points can be constructed.
Taking the feature point as a center, selecting a square pixel region with a certain size, dividing the region into 16 sub-regions with the same size, counting gradients of pixel points in the sub-regions in 8 directions (east, south, west, north, southeast, northeast, southwest and northwest) in each sub-region, and weighting the gradients by Gaussian filtering. That is, each sub-region has 8 directional gradients, and there are 16 sub-regions, that is, the descriptor of each feature point is a 128-dimensional vector. Since the rotation angle of the objective lens does not exceed 10 ° when taking a fluorescent picture, the influence of the rotation invariance on the feature point descriptors is not taken into account.
Because the adjacent two pictures are partially overlapped when the fluorescent picture is shot, the adjacent two pictures have a plurality of same characteristic points, and the adjacent two pictures can be spliced by matching descriptors of the same characteristic points.
According to the method, the nine pictures are spliced together at one time to form a panoramic picture of a slice of a certain layer of SBS modified asphalt.
And 4, constructing and optimizing the three-dimensional model.
By the method, panoramic pictures of all slices of the SBS modified asphalt test piece from bottom to top can be obtained, and the slices are numbered sequentially according to 1-N. On the basis, the panoramic image is converted from a color picture into a gray picture and then converted into a binary image, as shown in fig. 8, wherein the black part is asphalt, and the white part is the distribution of the SBS modifier and the asphalt light component after swelling.
Putting all binary images into a folder, and sequentially importing the binary images into Mimics software in batches, wherein the Mimics software can automatically sort the binary images according to picture numbers, the resolution of the process is set to be 90 mu m, and the slice interval is set to be 1 mu m so as to correspond to the slice experiment distance. Since the picture is subjected to binary processing by MATLAB in advance, complex binary conversion and various morphological operations on the picture in the Mimics software are not required. The Mimics software is skeleton CT tomogram three-dimensional structure reconstruction software commonly used in medical profession, and can extract pixels of pictures and place the pixels on a mask so as to quickly construct an original three-dimensional model.
Because the microstructure of SBS modified asphalt is highly irregular, the triangular patches used to form its shell are huge in number and should be cut down appropriately. This can be handled by the 3-matic software accompanying the mics software. And (3) introducing the SBS modified asphalt original three-dimensional model constructed in the Mimics software into 3-matic sub-software, and outputting the optimized three-dimensional model in the STL file type after reducing the number of triangular patches of the model and subdividing the three-dimensional model surface mesh.
And 5, materializing the three-dimensional model and importing finite elements.
However, the reconstructed model of the triangular patch by the Mimics software only has a shell and is not a solid model, that is, the model does not have any SBS modified asphalt information except the shape, so the model needs to be materialized.
Reading the STL file by SoildWorks software, popping up a dialog box by the SoildWorks software in the input process, and materializing the triangular patch model as long as the entity input is selected. Meanwhile, the SoildWorks software can also diagnose the defects of the three-dimensional model, repair a series of defects such as cracks and the like, and further improve the three-dimensional model; outputting the three-dimensional SBS modified asphalt model after the materialization in an x _ t file format, and then directly importing the three-dimensional SBS modified asphalt model into finite element software.
Examples
Step 1, preparing asphalt slices by quick freezing.
More than 0.1kg of SBS modified asphalt sample is heated in an oven at 180 ℃ for 2-3 hours at 170 ℃ and rapidly stirred for 30 seconds by using a glass rod, so that the sample is heated approximately uniformly. A 15mm × 15mm × 22mm glass rectangular mold with four sides is made, and placed on a glass plate, and a layer of spacer fluid (prepared from 50 Wt.% of talcum powder and 50 Wt.% of glycerol) is coated on the inner surface of the glass plate and the mold. Pouring the SBS modified asphalt which is in a liquid state after being heated into a mould by guiding a glass rod, and quickly scraping the surface by using a blade with the thickness of less than 0.5 mm. Three steel needles with the radius of 0.1mm are inserted into three corner positions of the slice, and the glass plate and the mould poured with the asphalt are placed in a refrigerator at the temperature of-10 ℃ for standby.
And 2, shooting the modified asphalt slice and the tomography fluorescence picture.
FIG. 5 is a schematic view of the tomographic scanning procedure of SBS modified asphalt slice. And taking out the frozen SBS modified asphalt sample, and slightly drawing out the positioning steel needle. The solid modified asphalt sample was cut into 1mm thick square sections in one direction using an ultrasonic rubber cutter. And removing the uppermost section and the lowermost section, placing the middle 20 sections at the cross-shaped position of the groove center of the glass slide, marking the three positioning holes of the first section sample on the glass slide, placing the subsequent samples at the cross-shaped center position, and then using tweezers to coincide the three positioning holes with the first sample positioning hole. And (4) slightly covering the cover glass and removing air bubbles between the cover glass and the modified asphalt. 20 samples are placed on a fluorescence microscopic objective table one by one, blue light (excitation light wavelength is 420nm-485nm) is turned on, 100-time amplification is adopted, and a focusing hand wheel is rotated to adjust the height of the objective table to observe a clear internal structure of the sample. After taking a picture from the central position, moving the longitudinal and transverse scales of the objective table to obtain other 8 fluorescence pictures at the position of the nine-square grid adjacent to the first picture, and taking pictures from the 1-9 nine-square grid mode according to the picture 6. Note that: the center of the image numbered 1 in the squared figure should coincide with the center of the slice. That is, each slice consists of 9 pictures and the sample consists of 20 layers, which means that there are a total of 180 SBS modified asphalt fluorescence pictures.
And step 3, splicing the pictures.
The picture splicing method is edited by MATLAB software, in order to enable the picture splicing to be smoother, according to the number of nine pictures, the pictures 1, 2 and 6 are spliced into the picture 10, the pictures 3, 4 and 5 are spliced into the picture 11, the pictures 7, 8 and 9 are spliced into the picture 12, the pictures 10 and 11 are spliced into the picture 13, and finally the pictures 12 and 13 are spliced into the picture 14, wherein the picture 14 is the layer of sliced panoramic picture.
According to the thought, all the 20 layers of slice pictures are spliced, and the number of the complete picture obtained after splicing is 1-20 from bottom to top in sequence, the number of the slice fluorescent picture at the lowest layer is 1, and the number of the slice picture at the uppermost layer is 20. The nine pictures before splicing of a slice in a certain layer are shown in fig. 7(a), and the complete picture after splicing is shown in fig. 7 (b).
Step 4, building and optimizing a three-dimensional model,
and (3) calling an MATLAB software rgb2gray function to convert each complete picture of the slices into a gray picture, calling a graythresh function to convert the gray picture into a binary picture, storing all the pictures into a tif format, and placing the pictures into a folder. Opening the Mimics software, inputting a folder where the slice images are located, and enabling the Mimics to automatically identify the serial numbers of the files, sort the files up and down and quickly construct an original three-dimensional model of the SBS modified asphalt. In order to construct a real SBS modified asphalt original three-dimensional model, the resolution of the Mimics software is set to be 90 μm, and the slice spacing is set to be 1mm to correspond to the slice experiment distance. Since the picture is subjected to binary processing in advance, the picture does not need to be subjected to binarization in the Mimics software, and various morphological operations are not needed.
The original three-dimensional model constructed by the Mimics software is a shell consisting of triangular patches, which means that the interior of the model is hollow, and the triangular patches used for constructing the three-dimensional model are fine and numerous because the microstructure of the SBS modified asphalt is highly irregular. Model optimization is therefore required.
The Mimics accessory software 3-matic is software specially used for optimizing a Mimics three-dimensional model, is an accessory program of Mimics and can directly receive the Mimics three-dimensional model. And directly introducing the three-dimensional model constructed in the Mimics into the 3-matic, and adjusting parameters in a Quality Preserving Reduce triangle function under a Remesh menu bar of the 3-matic, so as to Reduce a triangular patch of the three-dimensional model. And adjusting parameters in the UniformRemesh under the Remesh menu bar, and subdividing the mesh into a larger mesh for further reducing the number of triangular patches of the three-dimensional model. After the number of triangle panels is appropriate, the three-dimensional model is output in STL format.
And 5, materializing the three-dimensional model and importing finite elements.
And (3) importing the STL file into SoildWorks software, popping up a dialog box by the SoildWorks software, materializing the SBS modified asphalt triangular patch model by selecting the entity import, exporting the model in an x _ t file form, and finally importing the materialized SBS modified asphalt three-dimensional model into finite element software.
The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims (7)

1. An SBS modified asphalt three-dimensional microstructure reconstruction method based on an intelligent recognition algorithm is characterized by comprising the following steps:
freezing to prepare asphalt slices, comprising:
preparing SBS modified asphalt, preheating to a flowing state, pouring into a prefabricated test mold, inserting a steel needle at any three corners of the test mold for positioning, and freezing an SBS modified asphalt test piece at the temperature of minus 10 ℃ to minus 5 ℃ to enable the modified asphalt test piece to be in a solid state;
SBS modified asphalt fluorescence picture shooting, including:
pulling out the positioning steel needle, taking out the SBS modified asphalt test piece, cutting the solid modified asphalt test piece into N +2 layers of slices with the thickness of 1mm by using an ultrasonic rubber cutting knife, removing the two layers of slices on the upper surface and the lower surface, and numbering the rest N slices from bottom to top according to the number of 1-N; sequentially placing all the slices on a slide glass of a fluorescence microscope objective table for positioning, continuously shooting nine pictures at adjacent positions and numbering the pictures for the slices of each layer from a selected initial position according to a nine-square grid arrangement mode under the condition of eliminating bubbles between the slide glass and asphalt, and ensuring that the two adjacent pictures have partial overlapping areas;
picture stitching, including:
selecting nine pictures of each layer of slices, and sequentially splicing the nine pictures according to the photographing sequence to splice the nine pictures into a complete picture; the splicing process comprises the steps of establishing a differential pyramid for a picture, determining feature points in the picture, setting a main direction for the feature points, and then constructing descriptors of the feature points; splicing two adjacent pictures by matching descriptors of the same characteristic point, so that nine pictures of each layer of slices are spliced into a panoramic picture;
and respectively converting the panoramic pictures of all the slices into gray level images and binary images, establishing a three-dimensional model, performing triangular patch quantity reduction on the established three-dimensional model, dividing a three-dimensional model surface grid again, materializing the triangular patch model in the three-dimensional model, diagnosing the defects and repairing the defects to obtain the final materialized SBS modified asphalt three-dimensional model.
2. The SBS modified asphalt three-dimensional microstructure reconstruction method based on the intelligent recognition algorithm as claimed in claim 1, wherein the establishing of the differential pyramid for the picture comprises:
establishing a characteristic pyramid for the sliced pictures, wherein each rectangle in the characteristic pyramid represents one picture, the pictures with the same size form a group, one picture in the group is one layer, and different layers in the same group of pictures are formed by convolution of convolution kernels with different scales in the first layer in each group of pictures; any group of pictures in the characteristic pyramid are obtained by downsampling a previous group of pictures;
the formula for calculating the group number O and the layer number S of the characteristic pyramid is as follows:
O＝[log2(min(M,N))]-3
S＝n+3
in the above formula, M, N represents the length and width of the original picture, min represents the minimum value, [ ] represents the rounding, and n represents the number of the feature point pictures to be extracted;
the differential pyramid is the difference between two adjacent layers of pictures in the same group of the characteristic pyramid, the pictures in each group of the differential pyramid are one layer less than the pictures in the corresponding group of the characteristic pyramid, and the first layer of convolution kernel scale sigma in the differential pyramid0And the value of the parameter k is calculated by adopting the following formula:
k＝21/n
σ0＝1.52
in the first group, the first layer uses σ0As convolution scale, k σ is used for the second layer0As a convolution scale, the nth layer is represented by kn-1σ0As a convolution scale; and a second group of first layers with knσ0As a convolution scale, k for the second layern+1σ0As the convolution scale, the convolution kernel scales of the respective groups of layers and so on.
3. The SBS modified asphalt three-dimensional microstructure reconstruction method based on the intelligent recognition algorithm as claimed in claim 1, wherein the determining the feature points in the picture comprises:
calculating partial derivatives in three directions of a differential pyramid, including a picture length direction x, a picture width direction y and a picture kernel convolution direction sigma, wherein a point with a partial derivative of 0 is an extreme point;
the obtained extreme points are possibly picture feature points, and in order to further determine the positions of the feature points, the extreme points need to be accurately positioned, which includes:
in the differential pyramid, the pixel where the extreme point is located should be larger or smaller than the rest of the pixel points in the three directions, and before accurate positioning, the picture should be thresholded; reading the pixel value of each pixel point of the picture, and considering that the point cannot be an extreme point when the following formula is not met:
abs(val)≥0.5×D/n
in the above formula, abs (val) is the absolute value of the pixel point, D is an empirical parameter, 0.04 is taken, and n is the number of the feature point images to be extracted;
after thresholding of the picture, partial derivatives in three directions (X, y, sigma) are solved for the whole view angle space to obtain pixel points with partial derivatives of 0, and the coordinate of the pixel points is set as X0(x0,y0,σ0)T(ii) a Further, the extreme point of the picture is precisely located by using the pixel value function f (X) of the picture at the extreme point X0(x0,y0,σ0)TThe part is unfolded to obtain;
for all the extreme points found, if
Figure FDA0002835090430000031
The point pair is considered to have low contrast and is not a characteristic point; and eliminating the extreme points with edge effect for the rest extreme points, and taking the rest extreme points as feature points.
4. The SBS modified asphalt three-dimensional microstructure reconstructing method based on intelligent identification algorithm as claimed in claim 3, wherein the image extreme point is precisely located by applying image pixel value function f (X) at extreme point X0(x0,y0,σ0)TIs obtained by unfolding and comprises the following steps:
let the coordinate of the extreme point in the view space be X0(x0,y0,σ0)TThe picture pixel value function f (x) is expanded at this point:
Figure FDA0002835090430000032
in the above formula, f is a picture pixel value function having three components of x, y, and σ, the superscript T denotes a transpose matrix,
Figure FDA0002835090430000033
calculating a partial derivative; converting the above equation into a vector representation:
Figure FDA0002835090430000034
wherein X ═ X, y, σ ]; the partial derivatives of f (X) are calculated to obtain:
Figure FDA0002835090430000035
if the partial derivative is 0, solving the above equation yields:
Figure FDA0002835090430000036
Figure FDA0002835090430000037
the extreme point is taken into f (X) to obtain the extreme value
Figure FDA0002835090430000038
Figure FDA0002835090430000041
5. The SBS modified asphalt three-dimensional microstructure reconstructing method based on intelligent recognition algorithm as claimed in claim 3, wherein said eliminating extreme points with edge effect comprises:
the curvature of the pixel value function f (x) is represented by a matrix H, which is expressed as follows:
Figure FDA0002835090430000042
wherein D isxx(x,y)、Dxy(x,y)、Dyx(x,y)、Dyy(x, y) are each
Figure FDA0002835090430000043
Figure FDA0002835090430000044
A matrix formed by the derivation results;
the eigenvalues are replaced by traces tr (H) and determinant det (H) of the matrix H; the H matrix has two eigenvalues, set as α and β, where α > β, and α ═ γ β, γ being a coefficient; the relationship of the characteristic values to Tr (H) and Det (H) is represented by the following formula:
Tr(H)＝Dxx(x,y)+Dyy(x,y)＝α+β
Det(H)＝Dxx(x,y)Dyy(x,y)-(Dxy(x,y))2＝αβ
if Det (H) < 0, i.e., α and β are off, then there is an edge effect at this point and it should be dropped; if Det (H) > 0, α and β have the same sign, and α > β, so γ > 1, the following formula is calculated:
Figure FDA0002835090430000045
if it is
Figure FDA0002835090430000051
Then the point is considered to have an edge effect and should be discarded; if it is
Figure FDA0002835090430000052
The point is considered to be the desired extreme point, i.e., the feature point.
6. The SBS modified asphalt three-dimensional microstructure reconstructing method based on intelligent recognition algorithm as claimed in claim 1, wherein said setting a main direction for the feature point comprises:
taking the feature point as the center of a circle, taking 1.5 times of the convolution scale sigma of the difference pyramid picture where the feature point is located or is closest to the feature point as the radius, counting the gradient direction and the gradient amplitude of all pixel points in the circle, and performing 1.5 sigma Gaussian filtering operation to weight the pixel points at different positions in the circle, and taking the direction with the maximum amplitude as the main direction of the feature point.
7. The SBS modified asphalt three-dimensional microstructure reconstruction method based on the intelligent recognition algorithm as claimed in claim 1, wherein the building of the descriptors of the feature points realizes the splicing of two adjacent pictures by matching the descriptors of the same feature points, including:
selecting a square pixel area by taking the characteristic point as a center, dividing the area into 16 subregions with the same size, counting gradients of pixel points in the subregions in 8 directions in each subregion, and weighting the gradients by Gaussian filtering; each subregion has 8 directional gradients and 16 subregions, and the descriptor of each feature point is a 128-dimensional vector;
by matching the descriptors of the same feature points, the splicing of two adjacent pictures can be realized.
CN202011467882.4A 2020-12-14 2020-12-14 SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm Pending CN112632661A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011467882.4A CN112632661A (en) 2020-12-14 2020-12-14 SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011467882.4A CN112632661A (en) 2020-12-14 2020-12-14 SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm

Publications (1)

Publication Number Publication Date
CN112632661A true CN112632661A (en) 2021-04-09

Family

ID=75312953

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011467882.4A Pending CN112632661A (en) 2020-12-14 2020-12-14 SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm

Country Status (1)

Country Link
CN (1) CN112632661A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114049252A (en) * 2021-09-27 2022-02-15 中国科学院自动化研究所 Scanning electron microscope three-dimensional image acquisition system and method for sequence slicing
WO2023134310A1 (en) * 2022-01-11 2023-07-20 大连理工大学 High-throughput microstructure characterization and reconstruction method for heterogeneous material

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104298971A (en) * 2014-09-28 2015-01-21 北京理工大学 Method for identifying objects in 3D point cloud data
CN106289777A (en) * 2016-08-01 2017-01-04 北京航空航天大学 A kind of multi-state rolling bearing performance appraisal procedure based on geometry tolerance
CN107451985A (en) * 2017-08-01 2017-12-08 中国农业大学 A kind of joining method of the micro- sequence image of mouse tongue section
CN108921939A (en) * 2018-07-04 2018-11-30 王斌 A kind of method for reconstructing three-dimensional scene based on picture
CN109087396A (en) * 2018-07-26 2018-12-25 东南大学 A kind of microscopical structure reconstructing method based on CT images of concrete pixel characteristic
CN109932529A (en) * 2019-03-08 2019-06-25 河海大学 A method of based on polymer dispersity in atomic force microscope detection SBS modified pitch
CN110992408A (en) * 2019-11-18 2020-04-10 迪普派斯医疗科技(山东)有限公司 Digital section processing method and system based on pathological microscope

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104298971A (en) * 2014-09-28 2015-01-21 北京理工大学 Method for identifying objects in 3D point cloud data
CN106289777A (en) * 2016-08-01 2017-01-04 北京航空航天大学 A kind of multi-state rolling bearing performance appraisal procedure based on geometry tolerance
CN107451985A (en) * 2017-08-01 2017-12-08 中国农业大学 A kind of joining method of the micro- sequence image of mouse tongue section
CN108921939A (en) * 2018-07-04 2018-11-30 王斌 A kind of method for reconstructing three-dimensional scene based on picture
CN109087396A (en) * 2018-07-26 2018-12-25 东南大学 A kind of microscopical structure reconstructing method based on CT images of concrete pixel characteristic
CN109932529A (en) * 2019-03-08 2019-06-25 河海大学 A method of based on polymer dispersity in atomic force microscope detection SBS modified pitch
CN110992408A (en) * 2019-11-18 2020-04-10 迪普派斯医疗科技(山东)有限公司 Digital section processing method and system based on pathological microscope

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
MAOPING RAN等: "Asphalt Pavement Texture 3D Reconstruction Based on Binocular Vision System with SIFT Algorithm", 《2017 INTERNATIONAL CONFERENCE ON SMART GRID AND ELECTRICAL AUTOMATION (ICSGEA)》 *
刘雄等: "SBS改性沥青微观图像的采集及处理", 《公路与汽运》 *
张敏: "基于多视域广角相机视频图像拼接技术研究", 《中国博士学位论文全文数据库 (信息科技辑)》 *
武琳琳: "基于煤岩学研究手段构建桦甸油页岩有机质的三维结构", 《煤质技术》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114049252A (en) * 2021-09-27 2022-02-15 中国科学院自动化研究所 Scanning electron microscope three-dimensional image acquisition system and method for sequence slicing
WO2023134310A1 (en) * 2022-01-11 2023-07-20 大连理工大学 High-throughput microstructure characterization and reconstruction method for heterogeneous material
US11915386B2 (en) 2022-01-11 2024-02-27 Dalian University Of Technology High-throughput microstructure characterization and reconstruction method of heterogeneous materials

Similar Documents

Publication Publication Date Title
CN112632661A (en) SBS modified asphalt three-dimensional microstructure reconstruction method based on intelligent recognition algorithm
CN110210486B (en) Sketch annotation information-based generation countermeasure transfer learning method
CN107665492A (en) Colon and rectum panorama numeral pathological image tissue segmentation methods based on depth network
CN107194872A (en) Remote sensed image super-resolution reconstruction method based on perception of content deep learning network
Khoshdeli et al. Detection of nuclei in H&E stained sections using convolutional neural networks
CN106127221B (en) Classification of Polarimetric SAR Image method based on polarization-textural characteristics and DPL
CN104880389A (en) Mixed crystal degree automatic measurement and fine classification method for steel crystal grains, and system thereof
CN108053398A (en) A kind of melanoma automatic testing method of semi-supervised feature learning
Ivanovaitė et al. All these fantastic cultures? Research history and regionalization in the Late Palaeolithic tanged point cultures of Eastern Europe
CN113658117B (en) Method for identifying and dividing aggregate boundary in asphalt mixture based on deep learning
CN112270651B (en) Image restoration method for generating countermeasure network based on multi-scale discrimination
CN108830788A (en) A kind of plain splice synthetic method of histotomy micro-image
Doškář et al. Aperiodic compression and reconstruction of real-world material systems based on Wang tiles
CN110288613B (en) Tissue pathology image segmentation method for ultrahigh pixels
CN111476794B (en) Cervical pathological tissue segmentation method based on UNET
CN104992183A (en) Method for automatic detection of substantial object in natural scene
CN109635789A (en) Based on intensity than the High Resolution SAR image classification method that is extracted with spatial structure characteristic
CN103093231B (en) The automatic identification of second-phase and quantitative analysis method in a kind of alloy
Jafarbiglo et al. Nuclear atypia grading in histopathological images of breast cancer using convolutional neural networks
CN113989126A (en) X-ray defect detection image data enhancement method for deep learning
CN105139361B (en) A kind of image super-resolution rebuilding method of the FPM algorithms based on nonlinear optimization
Xiang et al. A deep network architecture for image inpainting
Weaver Technological analysis of prehistoric Lower Mississippi ceramic materials: a preliminary report
Wojnar et al. Quantitative image analysis
You et al. Automated cell individualization and counting in cerebral microscopic images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210409

WD01 Invention patent application deemed withdrawn after publication