CN101799393A - Automatic quantitative evaluation method of microstructure character of particulate matters discharged by diesel engine - Google Patents

Abstract

The invention discloses an automatic quantitative evaluation method of a microstructure character of particulate matters discharged by a diesel engine, which carries out automatic quantitative evaluation on three microstructure character parameters of the particulate matters discharged by the diesel engine, wherein the three microstructure character parameters include the length of each crystallite carbon layer on crystallite carbon layers of the particulate matters discharged by the diesel engine, the vertical distance of two adjacent crystallite carbon layers in the microstructure of a particle and the ratio of crystallite size to straight-line distance between pixel points at two ends of crystallite carbon layer. The method comprises four steps of: pre-processing a sample, acquiring a particle microcosmic feature image, carrying out mathematical conversion on the acquired particle microcosmic feature image; and extracting and calculating microstructure character parameters of the particulate matters discharged by the diesel engine. An image Gabor filtering method, a partial threshold value method, an improved OPTA (Optimal Performance Theoretically Attainable) method and the like are adopted for extracting the three microstructure character parameters of the particulate matters discharged by the diesel engine. With the method, the microstructure character of the particle can be evaluated automatically, quickly and accurately, thereby adding a quick and automatic evaluation method on the effect of a particle discharge control technology applied to vehicles and combustion engines.

Description

Automatic quantitative evaluation method of microstructure character of particulate matters discharged by diesel engine

Technical field

The present invention relates to a kind of evaluation method to the microstructure character of particulate matters discharged by diesel engine parameter.

Background technology

The atmosphere suspended particulate substance that diameter is small is because suspension time is long, specific surface area is big, and as easy as rolling off a log absorption poisonous and harmful substance threatens very big to health.The diesel locomotive institute exhaust gas discharged of One's name is legion is one of important source of suspended particulate substance in the urban atmosphere, and therefore the research to small particle emission of diesel car has been subjected to paying close attention to widely.At present the formation mechanism of diesel vehicle particle particularly particulate self structure and physicochemical property at it generates, Changing Pattern in the evolution course has become internal-combustion engine engineering and the environmental protection field is paid close attention to focus.

The particulate that forms in the diesel engine combustion has extremely complicated microstructure characteristic, and the multiple physicochemical property of this microstructure characteristic and particulate itself has confidential relation.As, many researchers have been found that this class for diesel emission is the particulate of core with the carbon, its oxidation characteristic changes with the variation of internal microstructure, particularly those have the carbonaceous particulate of amorphous micromechanism, easier to be more oxidized than the carbonaceous particulate with regular graphitization micromechanism, its oxidation reaction energy of activation is lower.Therefore, adopt the variation of institute's microstructure character of particulate matters discharged parameter in the reasonable method automatic quantitative evaluation diesel engine combustion, not only can deepen the formation mechanism and the discharging rule of research diesel particulation thing, the more important effective control method that is to help to propose to form in the diesel engine combustion particulate matter.

Summary of the invention

The objective of the invention is to carry out mathematic(al) manipulation and handle, propose a kind of technical method of automatic extraction microstructure character of particulate matters quantitative information by carbonaceous particulate microscopic appearance image to diesel emission.

Below in conjunction with accompanying drawing technical scheme of the present invention is described: the present invention solves how accurately to extract particulate microstructure characteristic parameter.These particulate microstructure characteristic parameters mainly comprise crystallite dimension La, basal spacing d and curvature C.Wherein La is to be an important structure parameter in the particulate heterogeneous microstructure of core with the carbon, and it is defined as the length of every crystallite carbon-coating on the crystallite carbon-coating; The definition of d is the vertical range of adjacent two crystallite carbon-coatings in the particulate micromechanism; The crystallite dimension that C is defined as the crystallite carbon-coating is the ratio of air line distance between the pixel of crystallite carbon-coating two ends therewith.Accompanying drawing 1 is seen in the definition of above-mentioned characteristic parameter.Above-mentioned 3 particulate microstructure characteristic parameters are very big to the multiple physical-chemical performance impact of particulate, particularly they are main determining factors of particulate oxide stability and high-temperature stability, their accurate evaluation are helped to improve the effect of the control technology of diesel particulate emission.To diesel engine discharged particle microstructural parameter automatic quantitative evaluation method, comprise four big steps, and wherein most importantly step 3, i.e. particulate microscopic appearance image mathematics transform process method.

Its concrete technical method that the present invention proposes is finished by following steps:

1. sample pre-treatments

The particulate samples fine gtinding that 1g is collected from diesel exhaust gas is placed on 150ml ethylene dichloride (CH ₂Cl ₂) in extraction 24 hours, during use ultrasonic oscillator to quicken described extraction process.Particulate samples after the extraction is placed 10ml absolute ethyl alcohol (CH ₃CH ₂OH) in, use ultrasonic oscillator vibration 30 minutes, the particulate samples after the grinding is dispersed in the absolute ethyl alcohol, form steady suspension.Get a scattered particulate-alcohol suspension and place on the little grid of transmission electron microscope (TEM) nickel screen, treat that ethanol freely volatilizees after, particulate samples pre-treatment work is promptly finished.

2. particulate microscopic appearance image obtains

Utilize a particulate samples pattern that transmission electron microscope (TEM) is finished step 1 to observe, obtain particulate microscopic appearance electron micrograph.

3. particulate microscopic appearance image mathematics transform process method.This step has comprised 6 main processes:

3.1 picture specificationization

Particulate microscopic appearance image is in shooting process, because artificial factor may cause contrast difference, the tonal range of the microscopic appearance image that is obtained variant.For statistical property indexs such as the average gray of diverse microcosmic appearance image, variance, contrast are adjusted to a unified scope, make different source images have identical gray average and variance, need be to the source images processing of standardizing.

If (i j) is pixel (i, gray scale j) to I, wherein to be respectively in the described particulate microscopic appearance image with the pixel be horizontal stroke, the ordinate of unit for i and j, and M and N represent the gray average and the gray variance of source particulate feature image, N (i respectively, j) be normalization back point (i, gray scale j).Picture specificationization is defined as follows:

In the formula,

$M=\frac{1}{A*B}\underset{i=0}{\overset{A-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{B-1}{\mathrm{\Σ}}}I(i,j)---\left(b\right)$

$V=\frac{1}{A*B}\underset{i=0}{\overset{A-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{B-1}{\mathrm{\Σ}}}{(I(i,j)-M)}^2---\left(c\right)$

The horizontal stroke that A and B have for the particle sources image in the formula, vertical pixel quantity when selecting to handle the scope of image, should make A close with B as far as possible; M ₀, V ₀Be gradation of image average and the variance after the predefined normalization, both spans are the integer between 0～255.

3.2 directional diagram calculates

At first, the particulate image after the normalization is divided into size is the sub-piece of Q * Q, wherein the span of Q is 0.35～10nm, and for the sub-piece of area less than Q * Q, is then removed.The sub-piece that so just the whole particulate image of A * B size is divided into P * P non-overlapping copies, P wherein can calculate by following formula:

$P=\frac{\mathrm{Min}(A,B)\×S}{Q}---\left(d\right)$

S is per two pairing actual ranges of neighbor pixel in the feature image in the formula.

Then, adopt Sobel operator (see figure 2) to calculate each sub-piece pixel (u, the level v) and the Grad of vertical direction

With

As formula (e) with (f):

$\∂x(u,v)=N(u+1,v-1)+2N(u+1,v)+N(u+1,v+1)-N(u-1,v-1)-2N(u-1,v)-N(u-1,v+1)---\left(e\right)$

$\∂y(u,v)=N(u-1,v+1)+2N(u,v+1)+N(u+1,v+1)-N(u-1,v-1)-2N(u,v-1)-N(u+1,v-1)---\left(f\right)$

Utilize again formula (g)-(i) estimation center point (i, j) local direction of sub-piece:

$V_x(i,j)=\underset{u=i-\frac{P}{2}}{\overset{i+\frac{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HSA00000017713500012.png"\; state="\{\{state\}\}">P</figure-callout>}}{2}}{\mathrm{\&Sigma;}}}\underset{v=j-\frac{P}{2}}{\overset{j+\frac{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HSA00000017713500012.png"\; state="\{\{state\}\}">P</figure-callout>}}{2}}{\mathrm{\&Sigma;}}}\&PartialD;x(u,v)\&PartialD;y(u,v)---\left(g\right)$

$V_y(i,j)=\underset{u=i-\frac{P}{2}}{\overset{i+\frac{\mathrm{<figure-callout\; id="2"\; label="P"\; filenames="HSA00000017713500012.png"\; state="\{\{state\}\}">P</figure-callout>}}{2}}{\mathrm{\&Sigma;}}}\underset{v=j-\frac{P}{2}}{\overset{j+\frac{P}{2}}{\mathrm{\&Sigma;}}}({\&PartialD;x}^2(u,v)-{\&PartialD;y}^2(u,v))---\left(h\right)$

θ in the formula (i) is the local direction angle of the sub-piece in place.

3.3 image segmentation comprises: variance method, direction method and composite algorithm.

The complex method that employing is made of jointly variance method and direction method is cut apart above-mentioned particulate feature image.Its ultimate principle is as follows: if the variance of gray scale is very big in the gray-scale statistical characteristic in a certain zone of image, then this zone is corresponding to the foreground area of image; Otherwise background area corresponding to image.In addition, if having peak value in the direction histogram in a certain zone of image, show that then this zone is a foreground area, otherwise be the background area.

3.3.1 variance method

3.3.1.1 the particulate feature image is divided into the sub-piece of the P * P of non-overlapping copies according to the method in the step 3.2, each image subblock is handled respectively;

3.3.1.2 calculate the gray average and the variance of each image subblock according to formula (j), (k):

$\mathrm{Means}=\frac{1}{P\&times;P}\underset{i=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{P}{\mathrm{\&Sigma;}}}N(u,v)---\left(j\right)$

$\mathrm{Vars}=\frac{1}{P\&times;P}\underset{i=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{P}{\mathrm{\&Sigma;}}}{(N(u,v)-\mathrm{Means})}^2---\left(k\right)$

N in the formula (u, v) be sub-piece (k, l) in the picture point gray-scale value of the capable v of u row, Means is the gray average of image subblock, Vars is the variance of image subblock.

3.3.1.3,, be set and be the background area as Vars during less than the threshold values T1 (span is the positive integer between 0～255) of definition for each image subblock; Otherwise,, keep its gray-scale value as foreground area.

3.3.2 direction method

3.3.2.1 calculate the directional diagram of particulate feature image according to the method in the step 3.2;

3.3.2.2 according to the method calculated direction histogram in the step 3.2;

3.3.2.3 each image subblock is cut apart by following standard:

If when the peak value in the 1. described direction histogram surpassed the threshold values T2 (span is the positive integer between 0～255) of definition, then this zone was regarded as prospect;

If the peak value difference in the 2. described direction histogram is less than the threshold values T3 (span is the positive integer between 0～255) of definition, then this zone is defined as background;

If 3. described direction variance is greater than the threshold values T4 (span is the positive integer between 0～10000) of definition, then this zone is defined as prospect.

3.3.3 composite algorithm

The image that if A, B are respectively the particulate feature image through direction method and variance method to be obtained after cutting apart, the split image of C for obtaining with complex method, then:

3.4 figure image intensifying

Image enchancing method adopts the real part of Gabor wave filter, the Gabor wave filter of even symmetry just, and its functional form is as follows:

In the formula, x _φ=xcos φ+ysin φ (n)

y _φ＝-xsinφ+ycosφ (o)

φ is the direction of crystallite carbon-coating in the formula, perpendicular to the Gabor wave filter; F is the frequency of crystallite carbon-coating; δ _xAnd δ _yBe respectively the Gauss constant of Gabor wave filter on x axle and y axle.

From formula (m)～(o) as can be seen,, must determine φ, f, parameter in order to obtain good enhancing effect, and the Gauss constant δ of Gabor wave filter on x axle and y axle _xAnd δ _y

3.4.1 calculate the directional diagram of each image subblock according to directional diagram Calculation Method in step 3.2 and 3.3;

3.4.2 calculate the frequency of crystallite carbon-coating

3.4.2.1 according to the image after the 3.1 steps normalization set by step the method in 3.2 be divided into the sub-piece of the non-overlapping copies of Q * Q size; With the image subblock central point (i j) is the center, and sub-piece carbon-coating direction is a minor axis, makes a rectangle window that is of a size of 2Q * Q size, and by formula (p) calculates amplitude X[k in window]:

k＝0，1，…，2P-1 (p)

In the formula,

$u=i+(d-\frac{P}{2})\cos \left(O(i,j)\right)+(k-P)\sin \left(O(i,j)\right)---\left(q\right)$

$v=j+(d-\frac{P}{2})\sin \left(O(i,j)\right)+(P-k)\cos \left(O(i,j)\right)---\left(r\right)$

3.4.2.2 discrete signal X[k] formed the sine wave of a two dimension, from the X[k that obtains] find all maximum points, and calculate the mean distance of these maximum points, or the mean pixel that is called between maximum point is counted, be designated as T (i, j), then the frequency of crystallite carbon-coating can be expressed as F (i, j)=and 1/T (i, j).

If 3.4.2.3 X[k] there is not continuous peak value in the signal, frequency values just is made as-1 so, the expression idling frequency.

3.4.3 determine the Gauss constant δ of Gabor wave filter on x axle and y axle _xAnd δ _y:

With the Gabor filter applies in the normalization after image N (u, the carbonaceous particulate microscopic appearance image E after v) can being enhanced (i, j):

$E(i,j)=\underset{u=-\frac{W_g}{2}}{\overset{-\frac{{\mathrm{<figure-callout\; id="2"\; label="W\; g"\; filenames="HSA00000017713500012.png"\; state="\{\{state\}\}">W</figure-callout>}}_g}{2}}{\mathrm{\&Sigma;}}}\underset{v=-\frac{W_g}{2}}{\overset{-\frac{W_g}{2}}{\mathrm{\&Sigma;}}}h(u,v:O(i,j)),F(i,j)N(i-u,j-v)---\left(s\right)$

(s) N is the particulate feature image after standardizing in the formula, and O is that crystallite carbon-coating directional diagram, F are crystallite carbon-coating frequency, W _gBe the size of Gabor wave filter, W _gSpan is 0～100 integer.

3.5 binaryzation

3.5.1 the method in 3.2 is divided into the particulate feature image the sub-piece of P * P size set by step;

3.5.2 each sub-piece is carried out following processing:

(1) by formula the method for (k) is asked the average gray of each sub-piece;

(2) in the sub-piece of statistics more than or equal to T and smaller or equal to the number of pixels N of T _lAnd N _s, T=Means wherein;

(3) if | N _s-N _l|＜δ (the δ span is 0～100 integer), then T is by being asked threshold values; N else if _s＞N _l, T=T-1 then; And | N _s-N _l| 〉=δ and N _s≤ N _lThe time, T=T+1.

3.6 image thinning

3.6.1 structure 8 elimination templates as shown in Figure 4, wherein 1 represents the foreground point, 0 expression background dot, and * expression promptly can be the foreground point and can be background dot again;

3.6.2 structure six reservation templates as shown in Figure 5;

3.6.3 from the top left corner pixel point of particulate micro image, according to from left to right, from top to bottom order scans feature image;

3.6.4, extract 14 pixel fields, as shown in Figure 6 for a certain foreground pixel point (is example with P5), preceding 8 field pixels (P1, P2, P3 with P5, P4, P6, P7, P8, P9) eliminate templates and compare with shown in the patent accompanying drawing 48, if when these 8 field elements and 8 couplings eliminating in the template (pixel value in the template in 8 fields of all elements of non-" * " value and this element definition claims to mate when all equating), then remove P5 (being P5=0), otherwise P5 keeps;

If 3.6.5 this pixel is removed in described step 3.6.4, then 14 field pixels with this pixel compare with 6 reservation templates shown in the accompanying drawing 5 again, remove three kinds of reservation template situations shown in Figure 7 simultaneously; If with one of them coupling, then P5 keeps, otherwise just really deletion of P5;

3.6.6 all pixels in the described particulate microscopic appearance image are carried out iteration, are changed the position up to the neither one pixel value.Result behind the image thinning as shown in Figure 8.

4. microstructure character of particulate matters discharged by diesel engine Parameter Extraction and calculating

After carrying out the image transformation of particulate microscopic appearance and handle according to step 3, extract particulate microstructure characteristic digital picture; According to the definition of particulate microstructure characteristic parameter, calculate these 3 particulate microscopic appearance characteristic parameters of crystallite dimension, basal spacing and curvature automatically again.

Adopt said method can realize that the automatic ration of particulate microscopic appearance characteristic parameter extracts.

Description of drawings

The definition of Fig. 1 particulate microstructure characteristic parameter

Fig. 2 Sobel operator template synoptic diagram

Fig. 3 direction window choose synoptic diagram

Fig. 4 eliminates the template synoptic diagram

Fig. 5 keeps the template synoptic diagram

The field synoptic diagram that Fig. 6 extracts

Fig. 7 keeps three kinds of situation synoptic diagram that template is removed

Particulate microscopic appearance figure after Fig. 8 refinement

The original micro image of Fig. 9 embodiment 1

The pending image of Figure 10 embodiment 1 intercepting

Result after Figure 11 embodiment 1 Flame Image Process

The original micro image of Figure 12 embodiment 2

The pending image of Figure 13 embodiment 2 interceptings

Result after Figure 14 embodiment 2 Flame Image Process

The original micro image of Figure 15 embodiment 3

The pending image of Figure 16 embodiment 3 interceptings

Result after Figure 17 embodiment 3 Flame Image Process

Embodiment

Below by specific embodiment the present invention is described in detail, but the content that the present invention is contained is not limited to following embodiment.

Embodiment one

1. sample pre-treatments

The particulate samples fine gtinding that 1g is collected from CY6102 diesel engine (sampling operating mode: rotating speed 1000r/min, injection pressure 110Mp, equivalent fuel air ratio 0.41) exhaust was placed in the 150ml ethylene dichloride extraction 24 hours, during use ultrasonic oscillator to quicken its extraction process.Particulate samples after the extraction is placed the 10ml absolute ethyl alcohol, use ultrasonic oscillator vibration 30 minutes, particulate samples is dispersed in the absolute ethyl alcohol, form steady suspension.Get a scattered described particulate-alcohol suspension and place on the little grid of TEM nickel screen, treat that ethanol freely volatilizees after, the particulate samples pre-treatment is finished.

2. particulate microscopic appearance image obtains

Utilize a transmission electron microscope that the particulate pattern is observed, obtain particulate microscopic appearance electron micrograph, as shown in Figure 9.In particulate microscopic appearance electron micrograph, select the apparent in view part of particulate microstructure characteristic, extract the image of 237 * 262 pixel sizes, as shown in Figure 10, carry out particulate microstructure characteristic Parameter Extraction.

3. particulate microscopic appearance image mathematic(al) manipulation

3.1 picture specificationization

Picture specification parameter M ₀And V ₀Be set at 0 and 255 respectively, then according to formula (a)～(c) carry out picture specificationization.

3.2 directional diagram calculates

If Q=0.35nm, the area of each sub-piece are 0.1225nm ², calculate S=0.265mm, P=1.79 * 10 ⁴

Then, adopt the Sobel operator according to formula (e) and (f) to calculate each sub-piece pixel (u, the level v) and the Grad of vertical direction

With Utilize formula (g)～(h) to estimate the local direction of each sub-piece again.

3.3 image segmentation

Set T1=0, T2=255, T3=0, T4=255.Carry out image segmentation according to the composite algorithm that formula (j)～(l) is described.

3.4 figure image intensifying

Behind the directional diagram that obtains described each sub-piece of particulate microscopic appearance image, by formula (p)～(r) calculates amplitude X[k], and calculate the mean distance of these maximum points, promptly obtain the crystallite carbon-coating frequency F (i, j)=1/T (i, j).

Set W _gValue is 0; Adopt the Gabor wave filter, the particulate microscopic appearance image after being enhanced according to the method in the formula (s).

3.5 binaryzation

According to the sub-piece of Q=0.35nm split image, setting the δ value is 0, and each sub-piece is carried out following processing: by formula the method for (j) is asked the average gray of each sub-piece; Add up in the sub-piece more than or equal to T and smaller or equal to the number of pixels N of T _lAnd N _s, T=Means wherein; If | N _s-N _l|＜δ, then T is by being asked threshold values; N else if _s＞N _l, T=T-1 then; And | N _s-N _l| 〉=δ and N _s≤ N _lThe time, T=T+1.

3.6 image thinning

Adopt as shown in Figure 48 to eliminate six reservation templates shown in template and the accompanying drawing 5; From the top left corner pixel point of image, according to from left to right, from top to bottom order handles as described in step 3.6 each pixel in the feature image, the result behind embodiment 1 image thinning is as shown in Figure 11.

4 microcosmos structure characteristic calculation of parameter

As calculated, each microcosmos structure characteristic parameter of embodiment 1 is respectively: La=1.0662nm; D=0.392131nm; C=1.305869.

Embodiment two

1 sample pre-treatments

The particulate samples fine gtinding that 1g is collected from CY6102 diesel engine (sampling operating mode: rotating speed 1000r/min, injection pressure 110Mp, equivalent fuel air ratio 0.53) exhaust was placed in the 150ml ethylene dichloride extraction 24 hours, during use ultrasonic oscillator to quicken extraction process.Particulate samples after the extraction is placed the 10ml absolute ethyl alcohol, use ultrasonic oscillator vibration 30 minutes, particulate samples is dispersed in the absolute ethyl alcohol, form steady suspension.Get a scattered described particulate-alcohol suspension and place on the little grid of TEM nickel screen, treat that ethanol freely volatilizees after, the particulate samples pre-treatment is finished.

Obtaining of 2 particulate microscopic appearance images

Utilize a transmission electron microscope that the particulate pattern is observed, obtain particulate microscopic appearance electron micrograph, as shown in Figure 12.In particulate microscopic appearance electron micrograph, select the apparent in view part of particulate microstructure characteristic, extract the image of 253 * 242 pixel sizes, as shown in Figure 13, carry out particulate microstructure characteristic Parameter Extraction.

3 particulate microscopic appearance image mathematic(al) manipulations

3.1 picture specificationization

Picture specification parameter M ₀And V ₀Be set at 255 and 0 respectively, then according to formula (a)～(c) carry out picture specificationization.

3.2 directional diagram calculates

If Q=5nm, the area of each sub-piece are 25nm ², calculate S=0.265mm, P=2.54 * 10 ³

Then, adopt the Sobel operator according to formula (e) and (f) to calculate each sub-piece pixel (u, the level v) and the Grad of vertical direction

With Utilize formula (g)～(h) to estimate the local direction of each sub-piece again.

3.3 image segmentation

Set T1=255, T2=0, T3=255, T4=0.Carry out image segmentation according to the composite algorithm that formula (j)～(l) is described.

3.4 figure image intensifying

Behind the directional diagram that obtains each sub-piece of particulate microscopic appearance image, by formula (p)～(r) calculates amplitude X[k], and calculate the mean distance of these maximum points, promptly obtain the crystallite carbon-coating frequency F (i, j)=1/T (i, j).

Set W _gValue is 100; Adopt the Gabor wave filter, the particulate microscopic appearance image after being enhanced according to the method in the formula (s).

3.5 binaryzation

According to the sub-piece of Q=5nm split image, setting the δ value is 0, and each sub-piece is carried out following processing: by formula the method for (j) is asked the average gray of each sub-piece; Add up in the sub-piece more than or equal to T and smaller or equal to the number of pixels N of T _lAnd N _s, T=Means wherein; If | N _s-N _l|＜δ, then T is by being asked threshold values; N else if _s＞N _l, T=T-1 then; And | N _s-N _l| 〉=δ and N _s≤ N _lThe time, T=T+1.

3.6 image thinning

Adopt as shown in Figure 48 to eliminate six reservation templates shown in template and the accompanying drawing 5; From the top left corner pixel point of image, according to from left to right, from top to bottom order handles as described in step 3.6 each pixel in the feature image, the result behind embodiment 2 image thinnings is as shown in Figure 14.

4 microcosmos structure characteristic calculation of parameter

As calculated, each microcosmos structure characteristic parameter of embodiment two is respectively: La=1.04720nm; D=0.35414nm; C=1.50231.

Embodiment three

1 sample pre-treatments

The particulate samples fine gtinding that 1g is collected from CY6102 diesel engine (sampling operating mode: rotating speed 1200r/min, injection pressure 110Mp, equivalent fuel air ratio 0.41) exhaust was placed in the 150ml ethylene dichloride extraction 24 hours, during use ultrasonic oscillator to quicken extraction process.Particulate samples after the extraction is placed the 10ml absolute ethyl alcohol, use ultrasonic oscillator vibration 30 minutes, particulate samples is dispersed in the described absolute ethyl alcohol, form steady suspension.Get a scattered described particulate-alcohol suspension and place on the little grid of TEM nickel screen, treat that ethanol freely volatilizees after, the particulate samples pre-treatment is finished.

Obtaining of 2 particulate microscopic appearance images

Utilize a transmission electron microscope that the particulate pattern is observed, obtain particulate microscopic appearance electron micrograph, as shown in Figure 15.In particulate microscopic appearance electron micrograph, select the apparent in view part of particulate microstructure characteristic, extract the image of 579 * 552 pixel sizes, as shown in Figure 16, carry out particulate microstructure characteristic Parameter Extraction.

3 particulate microscopic appearance image mathematic(al) manipulations

3.1 picture specificationization

Picture specification parameter M ₀And V ₀Be set at 100 and 150 respectively, then according to formula (a)～(c) carry out picture specificationization.

3.2 directional diagram calculates

If Q=10nm, the area of each sub-piece are 100nm ², calculate S=0.265mm, P=1.46 * 10 ⁴

Then, adopt the Sobel operator according to formula (e) and (f) to calculate each sub-piece pixel (u, the level v) and the Grad of vertical direction

With

Utilize formula (g)～(h) to estimate the local direction of each sub-piece again.

3.3 image segmentation

Set T1=100, T2=100, T3=100, T4=1000.Carry out image segmentation according to the composite algorithm that formula (j)～(l) is described.

3.4 figure image intensifying

Behind the directional diagram that obtains each sub-piece of particulate microscopic appearance image, by formula (p)～(r) calculates amplitude X[k], and calculate the mean distance of these maximum points, promptly obtain the crystallite carbon-coating frequency F (i, j)=1/T (i, j).

Set W _gValue is 50; Adopt the Gabor wave filter, the particulate microscopic appearance image after being enhanced according to the method in the formula (s).

3.5 binaryzation

According to the sub-piece of Q=10nm split image, setting the δ value is 100, and each sub-piece is carried out following processing: by formula the method for (j) is asked the average gray of each sub-piece; Add up in the sub-piece more than or equal to T and smaller or equal to the number of pixels N of T _lAnd N _s, T=Means wherein; If | N _s-N _l|＜δ, then T is by being asked threshold values; N else if _s＞N _l, T=T-1 then; And | N _s-N _l| 〉=δ and N _s≤ N _lThe time, T=T+1.

3.6 image thinning

Adopt as shown in Figure 48 to eliminate six reservation templates shown in template and the accompanying drawing 5; From the top left corner pixel point of image, according to from left to right, from top to bottom order is to 3.6 described processing the set by step of each pixel in the feature image, the result behind embodiment three image thinnings is as shown in Figure 17.

4 microcosmos structure characteristic calculation of parameter

As calculated, each microcosmos structure characteristic parameter of embodiment three is respectively: La=1.963nm; D=0.362nm; C=1.231.

The present invention has adopted a series of mathematical algorithms such as comprising image Gabor filter method, local threshold method, improved OPTA method to particulate microstructure characteristic parameter---crystallite dimension, curvature, basal spacing extract according to the characteristics of particulate microscopic appearance image.Use this method and can estimate the particulate microstructure characteristic automatically, fast and accurately, thereby the wire examination method fast and automatically of a kind of particulate emission control technology effect on automobile and internal combustion engine is provided.

Claims (1)

1. automatic quantitative evaluation method of microstructure character of particulate matters discharged by diesel engine, crystallite dimension (La), basal spacing (d) and 3 microcosmos structure characteristic parameters of curvature (C) of it is characterized in that automatic quantitative evaluation diesel emission carbonaceous particle thing, wherein crystallite dimension (La) definition is the length of every crystallite carbon-coating on the crystallite carbon-coating; The definition of basal spacing (d) is the vertical range of adjacent two crystallite carbon-coatings in the particulate micromechanism; The definition of curvature (C) is the crystallite dimension ratio of air line distance between the pixel of crystallite carbon-coating two ends therewith of crystallite carbon-coating, and the process of automatic quantitative evaluation may further comprise the steps:

(1) sample pre-treatments

The particulate samples fine gtinding that 1g is collected from diesel exhaust gas was placed in the 150ml ethylene dichloride extraction 24 hours, use ultrasonic oscillator to quicken extraction process during this time, particulate samples after the extraction is placed the 10ml absolute ethyl alcohol, use ultrasonic oscillator vibration 30 minutes, particulate samples after the grinding is dispersed in the described absolute ethyl alcohol, form steady suspension, getting a scattered particulate-alcohol suspension places on the little grid of transmission electron microscope nickel screen, after treating that ethanol freely volatilizees, particulate samples pre-treatment work is finished;

(2) particulate microscopic appearance image obtains

Adopt a transmission electron microscope that the described particulate samples pattern of step (1) is observed, obtain particulate microscopic appearance electron micrograph;

(3) image to step (2) the particulate microscopic appearance that obtains carries out the mathematic(al) manipulation processing:

(3.1) picture specificationization

To the source images processing of standardizing: establish I (i, j) be pixel (i, j) gray scale, wherein to be respectively in the particulate microscopic appearance image with the pixel be horizontal stroke, the ordinate of unit for i and j, M and N represent the gray average and the gray variance of source particulate feature image respectively, and (i j) is normalization back point (i to N, j) gray scale, picture specificationization is defined as follows:

In the formula,

$M=\frac{1}{A*B}\underset{i=0}{\overset{A-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{B-1}{\mathrm{\&Sigma;}}}I(i,j)---\left(b\right)$

$V=\frac{1}{A*B}\underset{i=0}{\overset{A-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{B-1}{\mathrm{\&Sigma;}}}{(I(i,j)-M)}^2---\left(c\right)$

The horizontal stroke that A and B have for the particle sources image in the formula, vertical pixel quantity; M ₀, V ₀Be gradation of image average and the variance after the predefined normalization, both spans are the integer between 0～255.

(3.2) directional diagram calculates

Particulate image after the normalization is divided into size is the sub-piece of Q * Q, wherein the span of Q is 0.35～10nm, removed less than the sub-piece of Q * Q for area, the whole particulate image of A * B size is divided into the sub-piece of P * P non-overlapping copies, wherein P calculates by (d) formula:

$P=\frac{\mathrm{Min}(A,B)\&times;S}{Q}---\left(d\right)$

S is per two pairing actual ranges of neighbor pixel in the feature image in the formula;

Adopt the Sobel operator to calculate each sub-piece pixel (u, Grad of horizontal direction v)

Grad with vertical direction Suc as formula (e) with (f):

$\&PartialD;x(u,v)=N(u+1,v-1)+2N(u+1,v)+N(u+1,v+1)-N(u-1,v-1)-2N(u-1,v)-N(u-1,v+1)---\left(e\right)$

$\&PartialD;y(u,v)=N(u-1,v+1)+2N(u,v+1)+N(u+1,v+1)-N(u-1,v-1)-2N(u,v-1)-N(u+1,v-1)---\left(f\right)$

Utilize again formula (g), (h), (i) estimation center point (i, j) local direction of sub-piece:

$V_x(i,j)=\underset{u=i-\frac{P}{2}}{\overset{i+\frac{P}{2}}{\mathrm{\&Sigma;}}}\underset{v=j-\frac{P}{2}}{\overset{j+\frac{P}{2}}{\mathrm{\&Sigma;}}}\&PartialD;x(u,v)\&PartialD;y(u,v)---\left(g\right)$

$V_y(i,j)=\underset{u=i-\frac{P}{2}}{\overset{i+\frac{P}{2}}{\mathrm{\&Sigma;}}}\underset{v=j-\frac{P}{2}}{\overset{j+\frac{P}{2}}{\mathrm{\&Sigma;}}}(\&PartialD;x^2(u,v)-\&PartialD;y^2(u,v))---\left(h\right)$

$\left\{\begin{array}{cc}\mathrm{\&theta;}(i,j)=\frac{1}{2}{\tan }^{-1\left(\frac{V_x(i,j)}{V_y(i,j)}\right)}& \mathrm{if}\left(V_x(i,j)\right)\&NotEqual;0\\ \mathrm{\&theta;}(i,j)=\frac{\mathrm{\&pi;}}{2}& \mathrm{if}\left(V_x(i,j)\right)=0\end{array}\right.---\left(i\right)$

θ in the formula (i) is the local direction angle of the sub-piece in place;

(3.3) carry out image segmentation, comprising: variance method, direction method, composite algorithm;

(3.3.1) variance method

(3.3.1.1) the particulate feature image is divided into the sub-piece of the P * P of non-overlapping copies according to step (3.2), each image subblock is handled respectively;

(3.3.1.2) calculate the gray average and the variance of each image subblock according to (j), (k) formula:

$\mathrm{Means}=\frac{1}{P\&times;P}\underset{i=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{P}{\mathrm{\&Sigma;}}}N(u,v),---\left(j\right)$

$\mathrm{Vars}=\frac{1}{P\&times;P}\underset{i=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{P}{\mathrm{\&Sigma;}}}{(N(u,v)-\mathrm{Means})}^2---\left(k\right)$

N in the formula (u, v) be sub-piece (k, l) in the picture point gray-scale value of the capable v of u row, Means is the gray average of image subblock, Vars is the variance of image subblock;

(3.3.1.3) for each image subblock, as Vars during, be set and be the background area, otherwise, keep its gray-scale value as foreground area less than the threshold values T1 of definition, the T1 span is the positive integer between 0～255;

(3.3.2) direction method

(3.3.2.1) calculate the directional diagram and the direction histogram of particulate feature image according to step (3.2);

(3.3.2.2) each image subblock is cut apart by following standard:

If when the peak value in the 1. described direction histogram surpassed the threshold values T2 of definition, then this zone was regarded as prospect, the T2 span is the positive integer between 0～255;

If the peak value difference in the 2. described direction histogram is less than the threshold values T3 of definition, then to be defined as background T3 span be positive integer between 0～255 in this zone;

If 3. described direction variance is greater than the threshold values T4 of definition, then this zone is defined as prospect, and the T4 span is the positive integer between 0～10000;

(3.3.3) composite algorithm

The image that if A, B are respectively the particulate feature image through direction method and variance method to be obtained after cutting apart, the split image of C for obtaining with complex method, then:

(3.4) figure image intensifying

Adopt the Gabor wave filter of even symmetry, its functional form is as follows:

$h(x,y:\mathrm{\&phi;},f)=\exp \{-\frac{1}{2}[\frac{x_{\mathrm{\&phi;}}^2}{{\mathrm{\&delta;}}_x^2}+\frac{y_{\mathrm{\&phi;}}^2}{{\mathrm{\&delta;}}_y^2}\left]\right\}\cos \left(2\mathrm{\&pi;f}{x}_{\mathrm{\&phi;}}\right)---\left(m\right)$

In the formula, x _φ=xcos φ+ysin φ (n)

y _φ＝-xsinφ+ycosφ (o)

φ is the direction of crystallite carbon-coating in the formula, perpendicular to the Gabor wave filter; F is the frequency of crystallite carbon-coating; δ _xAnd δ _yBe respectively the Gauss constant of Gabor wave filter on x axle and y axle, the figure image intensifying promptly needs to determine φ, f parameter, and the Gauss constant δ of Gabor wave filter on x axle and y axle _xAnd δ _y

(3.4.1) calculate the directional diagram of each image subblock according to directional diagram Calculation Method in the described step (3.2);

(3.4.2) frequency of calculating crystallite carbon-coating

(3.4.2.1) according to step (3.1) picture specification processed images more set by step the method in (3.2) be divided into the sub-piece of the non-overlapping copies of P * P size; With the image subblock central point (i j) is the center, and sub-piece carbon-coating direction is a minor axis, makes a rectangle window that is of a size of 2P * P size, and by formula (p) calculates amplitude X[k in window];

$X\left[k\right]=\frac{1}{P}\underset{d=0}{\overset{P-1}{\mathrm{\&Sigma;}}}N(u,v)$ k＝0，1，…，2P-1 (p)

In the formula,

$u=i+(d-\frac{P}{2})\cos \left(O(i,j)\right)+(k-P)\sin \left(O(i,j)\right)---\left(q\right)$

$v=j+(d-\frac{P}{2})\sin \left(O(i,j)\right)+(P-k)\cos \left(O(i,j)\right)---\left(r\right)$

(3.4.2.2) discrete signal X[k] formed the sine wave of a two dimension, from the X[k that obtains] find all maximum points, and calculate the mean distance of these maximum points, or the mean pixel that is called between maximum point is counted, be designated as T (i, j), then the frequency of crystallite carbon-coating can be expressed as F (i, j)=and 1/T (i, j);

If (3.4.2.3) X[k] there is not continuous peak value in the signal, frequency values just is made as-1 so, the expression idling frequency;

(3.4.3) determine the Gauss constant δ of Gabor wave filter on x axle and y axle _xAnd δ _y,

With the Gabor filter applies in the normalization after image N (u, the carbonaceous particulate microscopic appearance image E after v) can being enhanced (i, j):

$E(i,j)=\underset{u=-\frac{W_g}{2}}{\overset{-\frac{W_g}{2}}{\mathrm{\&Sigma;}}}\underset{v=--\frac{W_g}{2}}{\overset{-\frac{W_g}{2}}{\mathrm{\&Sigma;}}}h(u,v:O(i,j)),F(i,j)N(i-u,j-v)---\left(s\right)$

(s) N is the particulate feature image after standardizing in the formula, and O is that crystallite carbon-coating directional diagram, F are crystallite carbon-coating frequency, W _gBe the size of Gabor wave filter, W _gSpan is 0～100;

(3.5) binaryzation

(3.5.1) the particulate feature image is divided into the sub-piece of P * P size by the method in the described step (3.2);

(3.5.2) each sub-piece is carried out following processing:

(3.5.2.1) by formula the method for (k) is asked the average gray of each sub-piece;

(3.5.2.2) in the sub-piece of statistics more than or equal to T and smaller or equal to the number of pixels N of T _iAnd N _s, T=Means wherein;

If (3.5.2.3) | N _s-N _l|＜δ (the δ span is 0～100), then T is by being asked threshold values; N else if _s＞N _l, T=T-1 then; And | N _s-N _l| 〉=δ and N _s≤ N _lThe time, T=T+1.

(3.6) image thinning

(3.6.1) 8 matrixes of structure and form 8 and eliminate templates, 1 expression foreground point wherein, 0 expression background dot, * expression promptly can be the foreground point and can be background dot again;

$\left[\begin{array}{ccc}0& 0& 0\\ X& 1& X\\ 1& 1& 1\end{array}\right]$ $\left[\begin{array}{ccc}0& X& 1\\ 0& 1& 1\\ 0& X& 1\end{array}\right]$ $\left[\begin{array}{ccc}1& 1& 1\\ X& 1& X\\ 0& 0& 0\end{array}\right]$ $\left[\begin{array}{ccc}1& X& 0\\ 1& 1& 0\\ 1& X& 0\end{array}\right]$

$\left[\begin{array}{ccc}X& 0& 0\\ 1& 1& 0\\ X& 1& X\end{array}\right]$ $\left[\begin{array}{ccc}0& 0& X\\ 0& 1& 1\\ X& 1& X\end{array}\right]$ $\left[\begin{array}{ccc}X& 1& X\\ 0& 1& 1\\ 0& 0& X\end{array}\right]$ $\left[\begin{array}{ccc}X& 1& X\\ 1& 1& 0\\ X& 0& 0\end{array}\right]$

(3.6.2) 6 matrixes of structure and form six and keep templates;

$\left[\begin{array}{cccc}X& 1& X& 0\\ 0& 1& 1& 0\\ X& 1& X& 0\\ X& X& X& X\end{array}\right]$ $\left[\begin{array}{cccc}X& X& 0& 0\\ 0& 1& 1& 0\\ X& X& 1& X\\ X& X& X& X\end{array}\right]$ $\left[\begin{array}{cccc}X& X& 0& X\\ 0& 1& 1& 0\\ X& X& 0& 0\\ X& X& X& X\end{array}\right]$

$\left[\begin{array}{cccc}X& 0& X& X\\ 1& 1& 1& X\\ X& 1& X& X\\ 0& 0& 0& X\end{array}\right]$ $\left[\begin{array}{cccc}X& 0& X& X\\ X& 1& X& X\\ 1& 1& 0& X\\ X& 0& 0& X\end{array}\right]$ $\left[\begin{array}{cccc}X& 0& X& X\\ X& 1& X& X\\ 0& 1& 1& X\\ 0& 0& X& X\end{array}\right]$

(3.6.3) from the top left corner pixel point of image, according to from left to right, from top to bottom order scans feature image;

(3.6.4) for a certain foreground pixel point P _{(i, j)}(it is the pairing horizontal stroke of unit, ordinate with the pixel count that i and j are respectively this pixel) extracts 14 pixel fields, shown in following matrix, with P _{(i, j)}8 field pixel (P _{(i-1, j-1)}, P _{(i-1, j)}, P _{(i-1, j+1)}, P _{(i, j-1)}, P _{(i, j+1)}, P _{(i+1, j)}, P _{(i+1, j)}, P _{(i+1, j+1)}) eliminate templates and compare with top 8, if when these 8 field elements and 8 couplings eliminating in the template (pixel value in the template in 8 fields of all elements of non-" * " value and this element definition claims to mate when all equating), then remove P _{(i, j)}(be P _{(i, j)}=0), otherwise, P _{(i, j)}Keep;

$\left[\begin{array}{cccc}{P}_{(i-1,j-1)}& P_{(i-1,j)}& P_{(i-1,j+1)}& P_{(i-1,j+2)}\\ P_{(i,j-1)}& P_{(i,j)}& P_{(i,j+1)}& P_{(i,j+2)}\\ P_{(i+1,j-1)}& P_{(i+1,j)}& P_{(i+1,j+1)}& P_{(i+1,j+2)}\\ P_{(i+2,j-1)}& P_{(i+2,j)}& P_{(i+2,j+1)}& X\end{array}\right]$

If (3.6.5) this pixel is removed in described step (3.6.4), then with P _{(i, j)}14 field pixels 6 in again and (3.6.2) keep templates and compare, three kinds shown in the matrix keep the template situations below removing simultaneously; If with one of them coupling, then P _{(i, j)}Keep, otherwise P _{(i, j)}Just really deletion;

$\left[\begin{array}{cccc}0& 0& 1& 1\\ 0& 1& 1& 1\\ 0& 1& 1& 1\\ 0& 0& 1& 1\end{array}\right]$ $\left[\begin{array}{cccc}0& 0& 0& 0\\ 0& 1& 1& 0\\ 1& 1& 1& 1\\ 1& 1& 1& 1\end{array}\right]$ $\left[\begin{array}{cccc}1& 0& 0& 0\\ 1& 1& 0& 0\\ 1& 1& 0& 0\\ 1& 0& 0& 0\end{array}\right]$

(3.6.6) all pixels in the described particulate microscopic appearance image are carried out iteration, be changed the position up to the neither one pixel value;

(4) microstructure character of particulate matters discharged by diesel engine Parameter Extraction and calculating

After carrying out the image transformation of particulate microscopic appearance and handle according to step (3), extract particulate microstructure characteristic digital picture; According to the definition of particulate microstructure characteristic parameter, calculate these 3 particulate microscopic appearance characteristic parameters of crystallite dimension (La), basal spacing (d) and curvature (C) automatically again.

Images