A New Statistical-Based Kurtosis Wavelet Energy Feature for Texture Recognition of SAR Images
Gholamreza Akbarizadeh
Abstract—In this paper,an ef?cient algorithm for texture recog-nition of synthetic aperture radar(SAR)images is developed based on wavelet transform as a feature extraction tool and support vec-tor machine(SVM)as a classi?er.SAR image segmentation is an important step in texture recognition of SAR images.SAR images cannot be segmented successfully by using traditional methods because of the existence of speckle noise in SAR images.The algorithm,proposed in this paper,extracts the texture feature by using wavelet transform;then,it forms a feature vector composed of kurtosis value of wavelet energy feature of SAR image.In the next step,segmentation of different textures is applied by using feature vector and level set function.At last,an SVM classi?er is designed and trained by using normalized feature vectors of each region texture.The testing sets of SAR images are segmented by this trained SVM.Experimental results on both agricultural and urban SAR images show that the proposed algorithm is effective for classi?cation of different textures in SAR images,and it is also insensitive to the intensity.
Index Terms—Fourth-order normalized cumulant,kurtosis wavelet energy(KWE),SAR image classi?cation,speckle,syn-thetic aperture radar(SAR).
I.I NTRODUCTION
I N MANY applications,such as the global monitoring for the
environment,recognizing and tracking special objects,map-ping the Earth’s resources,and developing military systems, it is often bene?cial to have an imaging system which is able to provide broad-area imaging at high resolutions and acquire images in inclement weather or during night as well as day. Synthetic aperture radar(SAR)imaging system can provide these requirements.SAR enhances optical imaging abilities because of the unique reactions of xerographic targets to radar frequencies and because of the minimum restrictions on time of day and atmospheric situations.
SAR imaging systems are known as the most popular remote sensing technique greatly used in the past decades because of their capability to be utilized in all weather conditions,day and night photography time,and the high spatial resolution[1].In order to recognize and identify selected objects,SAR can pro-vide high-resolution images to distinguish terrain features[2]. However,SAR image processing is extremely dif?cult because
Manuscript received September29,2010;revised March27,2011,May14, 2011,July21,2011,and November10,2011;accepted April1,2012.Date of publication May23,2012;date of current version October24,2012.This work was supported by the Shahid Chamran University of Ahvaz as a research proposal with code901.
The author is with the Electrical Engineering Department,Engineering Faculty,Shahid Chamran University of Ahvaz,Ahvaz61357-831351,Iran (e-mail:g.akbari@scu.ac.ir).
Color versions of one or more of the?gures in this paper are available online at https://www.doczj.com/doc/976380975.html,.
Digital Object Identi?er10.1109/TGRS.2012.2194787of the speckle noise[2]–[5].The speckle noise is a fully developed noise which usually affects SAR images.Speckle phenomenon can be described as multiplicative noise,with standard deviation equal to pixel re?ectivity value[6].The probability density function(PDF)of the pixel intensities in SAR images is also impressed by speckle noise.This phe-nomenon can be expressed by the nonlinear intensity inhomo-geneity in SAR images[2].As a result of speckle noise effect on pixel intensities in SAR images,this is one of the main reasons for SAR imaging,which is a crucial issue for accurate segmentation and classi?cation.Accordingly,traditional seg-mentation and classi?cation methods based on intensity cannot be used to SAR image processing because of speckle noise effect on pixel intensity in SAR images.The main scope of the image segmentation and classi?cation is to categorize pixels into obvious image regions that are easier to analyze.Thus, segmentation can be used for image partitioning problems,and classi?cation can be used for object recognition purposes. Active contour models or snakes have been used as one of curve-evolution-based methods for global image segmen-tation[7].These methods are classi?ed into two major cate-gories:edge-based methods[8]–[10]and region-based methods [11]–[13].These methods either have weak performances in facing with weak object boundaries in SAR images,or they are sensitive to the location of initial contour and pixel intensities. Furthermore,these methods are only useful to segment the extended areas such as rivers and urban and agricultural areas. On the other hand,nonlinearity of intensity inhomogeneity,as mentioned earlier,often occurs in SAR images from different modalities.Intensity inhomogeneity can be addressed by some active contour models which are widely known as piecewise smooth models[14]–[17].Recently,Li et al.[18]have proposed a region-based active contour model by de?ning a region-scalable?tting energy function that locally approximates the image intensities on two sides of a contour.This model relies on the intensity inhomogeneity.Nevertheless,nonlinear intensity inhomogeneity which is usually attendant with SAR images cannot be addressed in all of these methods.
However,recent SAR image segmentation models have been developed containing generic segmentation procedure[19], [20],spectral data clustering algorithms[3],[21]–[23],fuzzy clustering algorithms[24],and level set methods[2],[4].
A generic segmentation method,which is conformed to gamma PDF of SAR images,is proposed by Galland et al.
[19]for SAR image segmentation.This parametric approach is on the basis of a polygonal grid model with a hypothetical unknown number of regions.In this approach,the number of regions of the partition is approximated by minimizing the stochastic complexity.However,when the gray values of SAR images are not correctly described by gamma PDF,like in
0196-2892/$31.00?2012IEEE
textured speckle images,this technique will fail.Furthermore, since a parametric noise reconstruction of the segmentation procedure is regarded,the parameters of regions need to be adjusted,particularly when the data differ from gamma PDF such as in textured regions.Thus,this parametric procedure only illustrates a unique class of textured data which can lead to analogous limitations to those obtained with the gamma PDF. The same gamma PDF distributed-based method as that in[19]was exerted with different noise models by Delyon and Réfrégier[20].As a comparison with the segmentation algorithm proposed in[19],this approach has some evident advantages.It is established on a polygonal grid which can have an arbitrary structure,and its region number and normalcy of its borders are acquired by minimizing the stochastic com-plexity of a determined quantity version on Q levels of the image[20].Unlike in[19],this approach reaches to a standard model without parameters that need to be tuned by the user. However,the proper value of Q,which minimizes the number of misclassi?ed pixels,could not be computed automatically in that procedure.Also,this procedure may fail due to the nonlin-ear intensity inhomogeneity phenomenon,which is mentioned earlier,if a texture region of a SAR image has other attributes of parametric noise models such as Poisson or Gaussian.
In[3]and[21]–[23],various schemes of spectral clustering algorithms were developed.These spectral clustering algo-rithms have very evident advantages compared with the tradi-tional clustering algorithms.Some of these spectral clustering algorithms can identify the clusters of irregular shapes and ob-tain the globally optimal solutions in a relaxed continuous do-main by eigendecomposition[3].However,the computational complexity problem is an imperfection that exists in these meth-ods,and they are computationally expensive because of the use of a coherence matrix?xed by the similarity of each pair of pix-els.Thus,the method needs to compute the eigenvectors of the coherence matrix.Furthermore,spectral clustering algorithms need to allocate a parameter,namely,the scaling parameterσin the Gaussian radial basis function(RBF).Appropriate allocat-ing ofσis a crucial issue to obtain good segmentation results in spectral clustering algorithms.Unfortunately,it is dif?cult to select the appropriateσvalue,and it is always set manually.The improper value ofσcan corrupt the abilities because spectral clustering algorithms are highly sensitive toσ,and different values ofσmay lead to extremely different results[3].
In[4]and[24],two new SAR image segmentation models were presented based on intensity homogeneity in each region as well as no purpose for segmenting the special objects in SAR images.On the other hand,nonlinearity of intensity inhomogeneity often occurs in SAR images as discussed earlier. These two new methods cannot address the nonlinear intensity inhomogeneity phenomenon.Moreover,these methods need an initial curve to be created by the user.
In this paper,we?rst propose an ef?cient method of SAR image segmentation by de?ning a new energy function,which is based on fourth-order normalized cumulant concept,named kurtosis wavelet energy(KWE).We demonstrate that KWE energy function can be used as an ef?cient feature for texture discrimination in SAR images.In other words,this statistic-based energy function is a good texture feature for SAR image segmentation problem.Kurtosis is a fourth-order normalized cumulant concluded by working toward the higher order statis-tics(HOS)in statistics subjects.We derive the KWE energy function by implementing the wavelet coef?cient energy extrac-tion algorithm and level set functions.Then,an SVM classi?er is designed and trained by using normalized feature vector composed of wavelet energy feature,KWE feature,and gray values of eight neighborhood of SAR image.This normalized feature vector is formed for each region texture of SAR image. Statistical properties of texture of each region in SAR image can be extracted well by this normalized feature vector because of using the higher order cumulant(fourth order)in feature formulation design.We will show in this paper that,whenever the order of cumulant as a feature for a SAR image increases, this feature will give the more statistical properties of a speci?c region from a SAR image,and subsequently,it will outperform the accuracy of texture recognition process of SAR image. Note that our feature extraction section method,termed as KWE,is also related to the skewness wavelet energy(SWE) model which is proposed in[2]where the skewness value of the wavelet coef?cient energy of the local intensity values in each region of SAR image is derived and the SWE values of the whole regions are used as the minimizers of the wavelet energy function.In[2],the lower order of cumulant(third order)is used as well as there is not any classi?er scheme,and only the segmentation process is done.On the other hand,in this paper, a local texture of each region of a SAR image is segmented well,and then,a classi?er is developed for texture recognition purposes.
This paper is organized as follows.In Section II,the feature extraction step for SAR image segmentation is performed by computing the KWE formulation as an effective feature for texture segmentation and classi?cation in SAR images.In Section III,derivation of the level set formulation with KWE energy is presented.In Section IV,the recognition step of our algorithm is performed.For this purpose,an SVM classi?er is designed and trained by using the KWE extracted feature for texture discrimination.The implementation results of our method on both agricultural and urban SAR images are given in Section V.Finally,Section VI draws some conclusions.
II.C UMULANTS AND K URTOSIS V ALUE AS A
D EFINITION FOR SAR T EXTURES
In statistics and probability theory,it can be shown that the cumulant generating function of a random variable X is expressed by the natural logarithm of the moment generating function of a random variable X as follows:
ΨX(ω)=ln(ΦX(ω))=ln
E X{e jωX}
(1)
in whichΦX(ω)is the moment generating function of a random variable X,E X{.}represents the mathematical expectation of that random variable,andωis the frequency variable of the Fourier transform.The moment generating functionΦX(ω)is also given by the Fourier transform of the PDF of the random variable X as follows:
ΦX(ω)=E X{e jωX}=
+∞
?∞
e jωx.
f X(x)dx.(2)
From the Fourier transform properties,it is obvious that the moment generating function takes the same information regarding the principal random variable as does the PDF.The natural logarithm functionΨX(ω)de?ned in(1)is usually mentioned as the cumulant generating function,and it is widely used in HOS.The cumulant generating function can be denoted by the cutoff Taylor series expansion as follows:
ΨX(ω)=
n
k=1
c X(k).
(jω)k
k!
(3)
where c X(k)is the coef?cient of the Taylor series and it is called the k th-order cumulant.It can be shown that the impression of cumulants gives a speci?cally powerful means for characterizing the nature of textures in images as stationary random series.Thus,it seems that cumulants can be good features to the description of textures.
It is reported that,whenever the order of cumulant as a feature for a SAR image increases,this feature will give the more statistical characteristics of a speci?c region from a SAR image[2].However,implementation of higher order cumulants will impose more calculations and is time consuming.Thus, we should perform a tradeoff between the higher order and the complexity of calculations.In[2],the third-order normalized cumulant named skewness was proposed to be used as a texture feature for segmentation of SAR images.In this paper,we propose to use the fourth-order normalized cumulant named kurtosis as a texture feature for segmentation step of our proposed algorithm.We show that this selection gives more statistical information of each region,and it has also more ef?cient implementation than the skewness.
It is feasible to represent the cumulants as functions of the moments of the random variable under analysis by using standard differentiation and mathematical identities.Note that the?rst-order moment is the mean moment,and the second-order moment is the central moment.For example,the?rst four cumulants as functions of the moments can be expressed by c(1)=m(1)=M=mean
c(2)=m(2)?[m(1)]2=σ2=covariance
c(3)=m(3)?3m(1)·m(2)+2[m(1)]3
c(4)=m(4)?3[m(2)]2?4m(1)·m(3)
+12[m(1)]2·m(2)?6[m(1)]4(4) where M andσ2represent the mean and covariance of the distribution of the related random variables,respectively.In order to clarify,the subscript X has been eliminated in these expressions.In most practical applications,the PDF of a ran-dom variable is unknown,and the cumulants must be computed from several comprehensions of the random variable.
The?rst step of our approach is segmenting different textures in SAR images using the kurtosis value of wavelet coef?cients of texture of each region as a texture feature.The formulation of third-and fourth-order cumulants c(3)and c(4),given by (3)and(4),respectively,can be used as such texture feature. In applications developed for segmentation of SAR images, a bene?cial feature is the feature that it is invariant to image size[2].However,the cumulant formulation as described in(4)is essentially dependent on the size of the generating random variable and also on the image size.For the purpose of this paper,we use a normalization of the higher order cumulants to make them invariant to image size.With this point of view, it is interesting to consider the normalized third-order and the normalized fourth-order cumulants with a desired random variable as de?ned respectively by
NC(3)=
c(3)
[c(2)]32
=
c(3)
[σ2]32
=
c(3)
σ3
(5)
NC(4)=
c(4)
[c(2)]2
=
c(4)
σ4
.(6)
Using this de?nition of cumulant,it directly follows that this normalized cumulant as a texture feature is image size invariant for any nonzero size of a SAR image.In(5),NC(3)is named skewness which is the third-order cumulant normalized with covariance.In(6),NC(4)is named kurtosis which is the fourth-order cumulant normalized with covariance.The skewness cumulant indicates the value of symmetry of the PDF histogram of a SAR image,and the kurtosis cumulant represents the sharpness of the PDF histogram of a SAR image. In other words,the kurtosis as a feature is the slope decreasing value of the PDF histogram curve of a SAR image.
It can be found that,whenever the order of cumulant as a feature for a SAR image increases,this feature will give the more statistical characteristics of a speci?c region from a SAR image[2].However,implementation of higher order cumulants will impose more calculations and is time consuming.Thus, we should perform a tradeoff between the higher order and the complexity of calculations.In[2],the third-order normalized cumulant named skewness is proposed to be used as a texture feature for segmentation of SAR images.In this paper,a texture representation of each region in SAR images with wavelet transform and kurtosis value concepts is de?ned.Our goal is to design and extract an ef?cient feature for texture segmentation of SAR images.For this purpose,we?rst apply a wavelet transform up to the possible last level on a SAR image,get the wavelet coef?cients,and compute the energy of the resulted wavelet coef?cients.Then,the kurtosis value of the wavelet coef?cient energy is calculated by computing the?rst four order moments of the wavelet coef?cient energies m(1)–m(4)and the four-order cumulant c(4)and at last applying(6)to get the value of the kurtosis NC(4).
In order to characterize a texture of each region in a SAR image by the wavelet coef?cient energy,we consider RT(i,j) as a region texture of a SAR image as follows:
RT(i,j)=
1
√
MN
m
n
W Aφ(y0,m,n)φy
,m,n
(i,j) +
x=H,V,D
?1
y=?y0
m
n
W xψ(y,m,n)ψx y,m,n(i,j)
?
?(7)
where?is the scaling function,ψis the wavelet function,y0is the order of the decomposition,W Aφis approximation wavelet
coef?cients,and W H
φ
,W V
φ
,and W D
φ
represent detail wavelet
coef?cients.The W H
φ
,W V
φ
,and W D
φ
wavelet coef?cients
are measured along the different directions as follows:W H
φalong columns(horizontal direction),W V
φ
along rows(vertical
direction),and W D
φalong diagonals.
We have tested several kinds of mother wavelets and several
wavelet coef?cients W A
φ,W H
φ
,W V
φ
,and W D
φ
.After exam-
ining the results,we have selected to use the Haar wavelet as mother wavelet type and the energy of approximation wavelet coef?cients in our proposed method.The energy of the approx-imation wavelet coef?cients in each subband can be obtained by the following sequence:
W A
φ(y,m,n)
2
,?y0≤y≤?1,m,n∈Z
.(8)
Aujol et https://www.doczj.com/doc/976380975.html,puted experimentally(see[26])that the dis-tribution of the square of the wavelet coef?cients in a subband of any image(also for SAR images)follows a generalized Gaussian law of the form
P X2(y)=
K
2
√
y
exp
?
√
y
α
β
y≥0
(9)
where K,α,andβare the texture parameters.IfΓ(t)and N represent the Gamma function and the total number of pixels of the given SAR image,respectively,then we have
K=
Nβ
αΓ
1
β
.(10)
To de?ne a texture feature based on kurtosis,we should
compute the fourth-order moment of the wavelet coef?cient energy distribution m(4)|W
φ|2
as follows:
m(4) |Wφ|2=E
|Wφ|8
=
+∞
(Wφ)8h(Wφ)d(Wφ)
=
Kα9
β
·Γ
9
β
.(11)
https://www.doczj.com/doc/976380975.html,parison between the curves of F?1(kurtosis),F?1(skewness),
and F?1(second moment).
Now,we can obtainαas follows:
α=
m(4)
|Wφ|2
m(3)
|Wφ|2
Γ
7
β
Γ
9
β
.(12)
Also,from c(4)and NC(4)de?ned by(4)and(6),respec-
tively,NC(4)can be rewritten as(13),shown at the bottom of
the page.
Thus,the formulation of function F(x)=NC(4)is then
obtained by substitution of the?rst-to fourth-order moments,
namely,m(1)|W
φ|2
,m(2)|W
φ|2
,m(3)|W
φ|2
,and m(4)|W
φ|2
,with their corre-
sponding values given in(14),shown at the bottom of the page.
Then,the value of classi?cation parameterβis calculated by
β=F?1(kurtosis)=F?1
?
??C(4)|Wφ|2
C(2)
|Wφ|2
2
?
??.(15)
The curve of F?1(kurtosis)is shown in Fig.1.As shown in
Fig.1,the curve of F?1(kurtosis)is approximately stable over
a wide range of kurtosis.In order to do a comparison between
the three methods,second moment,skewness,and kurtosis,we
have depicted the?gure of these methods in one plot as shown
NC(4)=C(4)
|Wφ|2
(σ2)4/2
=
m(4)
|Wφ|2
?3
m(2)
|Wφ|2
2
+4m(1)
|Wφ|2
·m(3)|W
φ|2
+12
m(1)
|Wφ|2
2
·m(2)|W
φ|2
?6
m(1)
|Wφ|2
4
σ4
(13)
F(x)=Γ3
1
β
Γ
9
β
?3·N·Γ2
1
β
·Γ2
5
β
?4·N·Γ2
1
β
·Γ
3
β
·Γ
7
β
N·
Γ
1
β
·Γ
5
β
?NΓ2
3
β
2
+q
12·N2·Γ
1
β
Γ2
3
β
·Γ
5
β
?6·N3·Γ4
3
β
N·
Γ
1
β
·Γ
5
β
?NΓ2
3
β
2(14)
Fig.2.De?ned segmentation problem.
in Fig.1.As shown in Fig.1,the curve of F?1(kurtosis)is
more stable than the F?1(skewness)and F?1(second moment)
which were proposed in[2]and[26],respectively.The curve
of F?1(kurtosis)is approximately stable over a wide range of
kurtosis.From Fig.1,as it is expected,the value ofβis about
0.25in all the ranges of kurtosis.Thus,if kurtosis is selected
as a texture feature for SAR image segmentation,it will be
an ef?cient feature because of its stability.As it is expected,
the kurtosis feature extracts more statistical information due
to its HOS.This is exactly the same thing that is required
to face with texture regions in SAR images.Thus,a texture
discrimination of each region in SAR images can be done by
the KWE explained in this section.In the next section,we
apply this feature in a level set function and develop it for
segmentation of each region in SAR images.
III.C OMPUTING THE L EVEL S ET F ORMULATION
W ITH KWE E NERGY T ERM
In the previous section,we described a new texture feature
extraction method for texture segmentation in SAR images.
In this section,we apply this feature in a level set function
and develop it for segmentation of each region texture in
SAR images.
Suppose that r is an open subset of R2,K is the number
of segmented regions(number of r k),k is a parameter that
shows the region index,x is a pixel of region,R kl is the
interface between region r k and region r l,and the image is a
function considered as I:r→R.We de?ne the region Re k= {x∈r|x belongs to the region k}.This de?ned segmentation problem is shown in Fig.2.
We denote that,for all k=1,...,K,Re k is an open set r k
given by a Lipschitz function L k:r→R so that
??
?L k(x)>0if x∈r k
L k(x)=0if x∈R k
L k(x)<0if otherwise
(16)
where R k is the boundary of r k and L k is the signed distance function to R k.We can determine r k using the sign of L k and the Heaviside distribution function H.This function is approximated by
Hα(β)=?
?
?
1
2
1+β
α
+1
π
sinπβ
α
if|β|≤α
1ifβ>α
0ifβ
(17)
In the distributional sense,whenα→0,we have Hα→H.
L k(x)is introduced as level set function.If L k(x)is calculated
for any point x and we get the sign of L k(x),we can determine
if x is in the region r k or not.For example,if L k(x)>0,then
H(L k(x))=1;thus,x∈k.Let x∈r be an arbitrary pixel and
I(x):r→R1be a given vector of SAR image,where1is
the dimension of the vector I(x).Dimension of SAR images
corresponds to the dimension of gray-level images.We de?ne
the following function:
F KWE(L1,L2,...,L K,f1,f2)
=εKWE
x
(L1,L2,...,L K,f1,f2)+μρ(L)(18)
whereμis a positive constant,εKWE
x
is the KWE,f1and f2are
the functions that minimize theεKWE
x
,andρ(L)is the deviation
of the level set function L from a signed distance function.The
kurtosis value of the wavelet coef?cient energy is proposed to
be used as contour energyεKWE
p
in(18).
Now,for a given pixel x∈r,the KWEεKW E
p
is followed
by the distribution form of the kurtosis energy of the wavelet
coef?cients in a subband of each region.For each point x∈r,
the KWE energy proposed in this paper is
εKWE
x
(C,k1(x),k2(x))
=γ1
inside contour(C)
K(x?y)|I(y)?k1(x)|2dy
=γ2
outside contour(C)
K(x?y)|I(y)?k2(x)|2dy(19)
where C is a contour in the image region r,γ1andγ2are
two constant numbers,K is a kernel function with a kurtosis
property,and k1(x)and k2(x)are two functions that?t image
textures near the pixel x.We call the pixel x the center point of
the aforementioned equation,and we call the aforementioned
energy the KWE around the center point x.A bene?t kernel
function K(x)should be computed and used in the KWE
energy function derived in(19).We propose to use the kernel
function K(x)as the PDF.Thus,we have
K|W
φ|2
(x)=
K
2
√
x
exp
?
√
x
α
β
(20)
where the constant parameter K is obtained by(10)and the
segment parametersαandβare given by(12)and(15),
respectively.
IV.R ECOGNITION OF S EGMENTED T EXTURES IN
SAR I MAGES W ITH SVM C LASSIFIER
The classi?er plays an important role in image classi?cation
and recognition.After the segmentation of textures in SAR
images is done,the classi?cation of each texture in SAR image
should be achieved by using an ef?cient classi?er and suitable
texture features.The theory of support vector machine(SVM),
as a tool of pattern classi?cation and recognition,is based on
statistical learning theory and the principle of structural risk
minimization.
SVMs are a set of af?liated supervised learning methods
which analyze data and recognize patterns.SVM is used for
statistical classi?cation and regression analysis.Suppose that,in a set of training examples,each example is marked as belonging to one of two categories.An SVM training algorithm initiates a model that predicts whether a new example drops within one category or the other.
An SVM creates a hyperplane or a set of hyperplanes in a high or in?nite dimensional space which can be used for classi?cation,recognition,or other tasks.Intuitively,a good clustering is achieved by the hyperplane which has the largest distance to the nearest training data points of any class.This nearest data point is also called functional margin.In general,the larger margin leads to the lower generalization error of the classi?er.The SVM classi?er belongs to a family of generalized linear classi?ers,but there are some ways to create nonlinear SVM classi?er by applying the kernel trick to maximum margin hyperplane.A linear SVM uses a systematic approach to ?nd a linear function with the lowest vapnik–chervonenis dimension.For nonlinear separable data,the SVM can map the input to a high-dimensional feature space where a linear hyperplane can be found.Thus,a good generalization can be attained by the SVM compared to traditional classi?ers.
The kernel function in a linear SVM is just a simple dot product in the input space.However,for a nonlinear SVM,the training data can be mapped to a feature space of higher dimension via a nonlinear projecting function.Since our clas-si?cation problem is a nonlinear form,hence,the optimal decision function can be considered as
f (x )=sgn m
i =1
a i y i K (x,x i )+
b (21)
where K (x,x i )is the kernel function.
In the SVM classi?er,the kernel function plays the important role of mapping the input samples into a feature space.At the present time,there is not any technique available to discover the structure of kernels.Typical choices of kernel function are the linear kernels,polynomial kernels,and Gaussian RBF ker-nels in SVM research.They are de?ned as follows:1)Linear kernels
K (x,x i )=x ?x i .
(22)
2)Homogeneous polynomial kernels
K (x i ,x j )=(x i ?x j )d .
(23)3)Inhomogeneous polynomial kernels
K (x i ,x j )=(x i ?x j +1)d .
(24)4)Gaussian RBF kernels
K (x i ,x j )=exp
?
x i ?x j 2
2σ2
.
(25)We use the RBF kernels as the kernel function and the arti?cial choice method to obtain the samples in our proposed method for classi?cation of segmented textures in SAR images.After the original SAR image was ?ltered by wavelet trans-form,the energy values of wavelet coef?cients,the kurtosis value of wavelet coef?cient energy,and gray values of eight neighborhood of that will be computed.These computed
values
Fig.3.(a)Three-look simulated SAR image (256×256).(b)Ground truth.(c)Segmentation obtained by LBF model (error rate:4.68%;the number of missegmented pixels:3070).(d)Segmentation obtained by SWE model (error rate:3.33%;the number of missegmented pixels:2185).(e)Segmentation obtained by KWE (error rate:1.46%;the number of missegmented pixels:954).
compose the feature vector of samples.Then,segmentation of textures is applied by using feature vector and level set function proposed in Section III.At last,an SVM classi?er is designed and trained by using normalized feature vectors of each region texture,and the testing sets of SAR image are divided by the trained SVM.According to the classi?cation results,the gray value whose category is “+1”was set at 255,and the gray value whose category is “?1”was set at 0.Thus,the segmentation of SAR image is realized.
V .I MPLEMENTATION AND T EST R ESULTS
To elucidate the relative advantages of the KWE with respect to local binary ?tting (LBF)and SWE,the results of different algorithms on simulated and real SAR images are presented.A.Segmentation of Simulated SAR Image
In order to evaluate the performance of the proposed method objectively,we ?rst show an experiment on a simulated
Fig.4.(a)X-SAR image of Washington,D.C.(512×512).(b)Aerial optical photograph of the same region as the one of the SAR image(adopted by Google Earth).(c)Representation of the approximation wavelet coef?cient energy|W A
φ
|2from levels1to9of the“Washington,D.C.,”SAR image.
three-look SAR image.The generation procedure of the sim-ulated SAR image was inspired by radar image formation phenomena.This is done by averaging three gamma-distributed realizations.The corresponding three-look noisy image,as shown in Fig.3(a),is generated by averaging three independent realizations of speckle.The ground truth image,as shown in Fig.3(b),is used to calculate the error rates of the segmenta-tions obtained by different algorithms.
Three algorithms are used for segmentation respectively: 1)the LBF model;2)SWE;and3)KWE.Fig.3(c)shows the segmentation result with the LBF model.Fig.3(d)shows the segmentation result of SWE,which is better than the LBF model.Fig.3(e)shows the best segmentation result according to the error rate.We found that the overall error rates are reduced from4.68%to3.33%by using SWE and to1.46% by using KWE.Therefore,SWE is better than LBF,and KWE outperforms the SWE.
Visually,the segmentation of the LBF,which is shown in Fig.3(c),is seriously spotty in consistent regions.Many pixels in two segments are confused.SWE performs better than the LBF,as shown in Fig.3(d).Therefore,SWE is more robust to the noise than the LBF.In the result,the missegmented pixels mainly locate in the white regions of Fig.3(a),and the black regions in Fig.3(a)are well https://www.doczj.com/doc/976380975.html,pared with SWE, KWE reduces the number of the missegmented pixels,as shown in Fig.3(e).
B.Segmentation of Real SAR Images
When we deal with real SAR image segmentation,the ground truth corresponding to the SAR images being seg-mented is absent generally.In this case,the evaluation of the segmentation result is based on visual inspection of the segmented
images.Fig.5.SVM classi?er with an RBF kernel which is trained by using the texture features of two different textures of the original“Washington,D.C.,”SAR
image.
Fig.6.Test stage of the trained SVM classi?er.Two different texture data sets of feature vectors of“Washington,D.C.,”SAR image is applied to the classi?er, and the clustered data are classi?ed in two different classes with violet and blue colors.
In this section,in order to verify the effect of the proposed
method,experiments on both agricultural and urban SAR im-ages are performed.
Fig.4(a)shows an original SAR image.It is a National Aeronautics and Space Administration(NASA)Goddard Space Flight Center image with15-m resolution of Washington,D.C., acquired by LANDSAT7on May9,2005.This image is an urban X-band SAR image whose size is512×512.Fig.4(b) shows an aerial optical photograph of the same region of the SAR image shown in Fig.4(a).In order to extract the texture features in different regions of SAR image,the wavelet transform is?rst applied to the image to get the approximation wavelet coef?cients W A
φ
.Then,the values of the energy of the wavelet coef?cients are computed by means of the expression obtained in(8).Note that,in order to extract all of the wavelet coef?cients,the wavelet decompositions of the maximum level L have been used to compute all of the approximation
wavelet coef?cients W A
φ
and then,the?rst-,second-,third-, and fourth-order moments of the energy distribution of these wavelet coef?cients are calculated.For example,in SAR image,
Fig.7.(a)Segmentation obtained by the LBF.(b)Segmentation obtained by SWE.(c)Segmentation obtained by the proposed KWE.(d)[respectively,(e)and (f)]Zoom of an area extracted from(a)[respectively,(b)and(c)].(g)Zoom of the same area extracted from Fig.5(b).
as shown in Fig.4(a),L=9(because512=29).Fig.4(c)
shows the image representation of the approximation wavelet
coef?cient energy|W A
φ|2from levels1to9of the“Washington,
D.C.,”SAR image.
Now,we can extract the KWE feature from the approxima-tion wavelet coef?cient energy shown in Fig.4(c)for X-SAR image of the“Washington,D.C.”The value of the kurtosis of the approximation wavelet coef?cient energy shown in Fig.4(c) is6.514×1012.It is obtained by(13).Now,we can get β=0.2598from the extended curve of Fig.1in the range [0,12×1012]of only the kurtosis axis in detail.
Also,we calculatedα=0.2552and K=2.1436×105 from(12)and(10),respectively.Note that the classi?cation parametersβ,α,and K are different for each SAR image. After extracting the texture feature,segmentation and classi-?cation algorithm of each texture should be carried out on SAR image with these extracted texture features.To implement our level set function with KWE texture feature,the parameters in the experimentation are supposed as follows:γ1=1.0,γ2= 1.0,ν=0.004×2552,c0=2(constant value of step function used as initial contour),time stepτ=0.1,μ=1.0,and center point p=1.0.Also,an SVM classi?er is designed and trained by using the normalized feature vector.In Fig.5,an SVM classi?er with an RBF kernel is designed and trained by using the feature vector of two different textures of the“Washington, D.C.,”SAR image.
As shown in Fig.5,the SVM classi?er is trained with two sets of feature vector.One,which is de?ned with+,is labeled with“1”and red color,and the other,which is de?ned with?, is labeled with“2”and green color.The class“1”is composed of the texture features of water regions,and the class“2”is composed of the texture features of vegetation regions of the “Washington,D.C.,”SAR image.This trained SVM classi?er is determined by a line which is obtained based on SVM principles and the nearest trained data which are labeled as support vectors.
After training with the proposed KWE feature subspace, the SVM classi?er is used to recognize the texture classes of water–vegetation subblocks of the input SAR image.For this purpose,another data set is used to test the trained SVM classi?er.The test stage of the SVM classi?er with two different texture data sets is shown in Fig.6.
In Fig.6,two different feature vectors of two different textures of the“Washington,D.C.,”SAR image are applied to the classi?er.The?rst cluster is the feature vectors of the texture of water regions which is classi?ed in the?rst class.This class is labeled as“1(classi?ed)”with violet color,and it is speci?ed with“+”signs.The second cluster of data sets is the feature vectors of the vegetation regions of the“Washington,D.C.,”SAR image which is labeled as“2(classi?ed)”with blue color, and it is speci?ed with“?”signs.
In the same way,the proposed SVM classi?er can be trained by the texture features of the building regions of the “Washington,D.C.,”SAR image.Now,we can use this SVM classi?er for segmentation and classi?cation of different tex-tures of SAR images.In order to examine the capability of the proposed method in this paper,two traditional methods such as SWE[2]and LBF[18]are selected to have some comparisons. The“Washington,D.C.,”SAR image as shown in Fig.4(a) is used?rst to segmentation and classi?cation operation.The experimental results of LBF model as a pixel-intensity-based method are shown in Fig.7(a).The experimental results of SWE model as a segmentation method based on skewness wavelet energy is shown in Fig.7(b).The results of the pro-posed method are shown in Fig.7(c).The magni?ed versions of the same selected area are shown in Fig.7(d)–(g).
The images,as shown in Fig.7,consist of three types of land cover:water,vegetation,and building.The water area is marked as region A,the vegetation area is marked as region B, and the building area is marked as region C in the results. These images are urban SAR images.The segmentation ob-tained by the LBF model is shown in Fig.7(a).One can see that the water area(lower left)is incorrectly segmented.In other words,the LBF model has failed while facing particular regions such as water.Furthermore,the boundary between the vegetation[region B in Fig.7(d)]and the building[region C in Fig.7(d)]is not correctly de?ned.The segmentation obtained by SWE,as shown in Fig.7(b),improves the uniformity in the water region.However,there is serious missegmentation in the vegetation region[see the magni?ed image of a se-lected area in Fig.7(e)].Furthermore,both LBF and SWE models need an initial contour created by the user.KWE gets the best segmentation and classi?cation result,as shown in Fig.7(c).
The classi?cation operation,obtained by KWE,improves the uniformity in the water region,and a local region of the region B [part of region B located in region C in Fig.7(f)]is consistently recognized as vegetation.Three types of land cover in Fig.7 are consistently identi?ed as corresponding regions by using the KWE classi?cation algorithm.Moreover,the boundaries of particular regions are well determined by KWE.Also,the proposed KWE classi?cation algorithm does not need an initial contour selected by the user.Thus,the KWE can be utilized in automatic processes.
Another experiment is carried out on an agricultural SAR image in Fig.8(a).This SAR image is a multilook C-band
SAR Fig.8.(a)C-SAR image of a rice-growing area near Okayama,Japan, obtained by JPL AirSAR(1024×1024).(b)Aerial optical photograph of the same region as the one of the SAR image(adopted by Google map).
(c)Representation of the approximation wavelet coef?cient energy|W A
φ
|2 from levels1to10of the“rice-growing”SAR image.
image of a rice-growing area near Okayama,Japan,obtained
by NASA/Jet Propulsion Laboratory AirSAR.This image is already multilooked nine times in azimuth to give a pixel spac-ing of approximately4.6m in azimuth and3.3m in range[25]. This image is an agricultural C-SAR image whose size is 1024×1024.Fig.8(b)shows an aerial optical photograph of the same region as the one of the SAR image.The approxi-
mation wavelet coef?cient energy|W A
φ
|2from levels1to10 of the“rice-growing”SAR image is also shown in Fig.8(c). This image consists of?ve types of land cover:water,urban, vegetation,rice,and wheat.
The segmentation obtained by the LBF method is shown in Fig.9(a),and a zoom of an area extracted from this?gure is shown in Fig.9(d).The water area on the down left is segmented badly,and one can see that two water local regions in the rice area[see the down middle image in Fig.9(d)]are mistakenly segmented with the vegetation,and a big part of the rice area is mistakenly segmented with the wheat area. Therefore,the LBF model is not effective for segmentation of this image.SWE improves the segmentation result to some degree,as shown in Fig.9(b).The magni?ed version of this image for the same area is shown in Fig.9(e).The uniformity in the water area is improved,and the vegetation area is identi?ed as well.However,the rice area is segmented badly,and the urban area is not identi?ed.The segmentation and classi?cation of the proposed KWE shows an effective classi?cation result in comparison with those of the LBF and SWE,as shown in Fig.9(c).The uniformity in the rice area and the water area is improved,and the urban area and the wheat area are identi?ed as well[see the magni?ed image as shown in Fig.9(f)].
VI.C ONCLUSION
We have developed a new segmentation and classi?cation algorithm based on kurtosis value of wavelet coef?cient energy for segmentation of each region and recognition of each texture
Fig.9.(a)Segmentation obtained by the LBF.(b)Segmentation obtained by SWE.(c)Segmentation and classi?cation obtained by the proposed KWE.
(d)[respectively,(e)and(f)]Zoom of an area extracted from(a)[respectively,(b)and(c)].(g)Zoom of the same area extracted from Fig.9(b).
of SAR images.A new energy named KWE is proposed to be used as a feature for texture discrimination of each region. In comparison with the LBF segmentation model,the KWE achieves better performance on the SAR images.It also per-forms better than the SWE in often cases because it extracts more statistical information of textures due to its higher order of cumulant.Furthermore,the KWE algorithm,proposed in this paper,avoids the selection of the initial contour by the user. Thus,the KWE algorithm can be used in automatic processes. Experimental results show that the proposed method is more ef?cient for accurate segmentation and classi?cation of several kinds of SAR images.
VII.F UTURE W ORK
In this paper,we have considered the kurtosis of the energy as the texture feature.The kurtosis is linked with the Fourier transform which is de?ned in R,but the energy is always a positive variable.A more appropriate transform to represent the energy will be the Mellin transform.In this case,the log cumulants are considered instead of classical moments.Thus, second kind statistics and the Mellin transform offer a better adapted formalism for positive random variables which could be a suitable topic for future work.Also,the reader can study the estimation methods such as Fisher information matrix and the Cramer–Rao bound to extract good features from SAR images in order to reach a better segmentation result.
A CKNOWLEDGMENT
The author would like to thank the Shahid Chamran Univer-sity of Ahvaz for?nancial support.
R EFERENCES
[1]G.A.Rezai-Rad and G.Akbarizadeh,“A new readout circuit structure for
SAR satellite imaging sensors,”Proc.IREE,vol.5,no.1,pp.281–290, Feb.2010.
[2]G.Akbarizadeh,G. A.Rezai-Rad,and S. B.Shokouhi,“A new
region-based active contour model with skewness wavelet energy for
segmentation of SAR images,”IEICE Trans.Inf.Syst.,vol.E93-D,no.7, pp.1690–1699,Jul.2010.
[3]X.Zhang,L.Jiao,F.Liu,L.Bo,and M.Gong,“Spectral clustering ensem-
ble applied to SAR image segmentation,”IEEE Trans.Geosci.Remote Sens.,vol.46,no.7,pp.2126–2136,Jul.2008.
[4]Y.Shuai,H.Sun,and G.Xu,“SAR image segmentation based on level
set with stationary global minimum,”IEEE Trans.Geosci.Remote Sens., vol.5,no.4,pp.644–648,Oct.2008.
[5]S.Li,Z.Yanning,M.Miao,and T.Guangjian,“SAR image segmentation
method using DP mixture models,”in Proc.IEEE https://www.doczj.com/doc/976380975.html,put.Sci.
Comput.Technol.,Dec.2008,vol.2,pp.598–601.
[6]L.R.Varshney,“Despeckling synthetic aperture radar imagery using the
contourlet transform,”2004.[Online].Available:https://www.doczj.com/doc/976380975.html,/~lrv/ www/cornell/publications/426%20Report%203.pdf
[7]C.Li,C.Kao,J.Gore,and Z.Ding,“Implicit active contours driven by
local binary?tting energy,”in Proc.IEEE CVPR,Jun.2007,pp.1–7. [8]C.Li,C.Xu,C.Gui,and M.D.Fox,“Level set evolution without
re-initialization:A new variational formulation,”in Proc.IEEE CVPR, Jun.2005,vol.1,pp.430–436.
[9]C.Xu and J.Prince,“Snakes,shapes,and gradient vector?ow,”IEEE
Trans.Image Process.,vol.7,no.3,pp.359–369,Mar.1998.
[10]V.Caselles,R.Kimmel,and G.Sapiro,“Geodesic active contours,”Int.J.
Comput.Vis.,vol.22,no.1,pp.61–79,Feb./Mar.1997.
[11]N.Paragios and R.Deriche,“Geodesic active regions and level set meth-
ods for supervised texture segmentation,”https://www.doczj.com/doc/976380975.html,put.Vis.,vol.46, no.3,pp.223–247,Feb./Mar.2002.
[12]C.Samson,L.Blanc-Feraud,G.Aubert,and J.Zerubia,“A variational
model for image classi?cation and restoration,”IEEE Trans.Pattern Anal.
Mach.Intell.,vol.22,no.5,pp.460–472,May2000.
[13]R.Ronfard,“Region-based strategies for active contour models,”Int.J.
Comput.Vis.,vol.13,no.2,pp.229–251,Oct.1994.
[14]O.Michailovich,Y.Rathi,and A.Tannenbaum,“Image segmentation
using active contours driven by the Bhattacharyya gradient?ow,”IEEE Trans.Image Process.,vol.16,no.11,pp.2787–2801,Nov.2007. [15]L.Vese and T.Chan,“A multiphase level set framework for image seg-
mentation using the Mumford and Shah model,”https://www.doczj.com/doc/976380975.html,put.Vis., vol.50,no.3,pp.271–293,Dec.2002.
[16]D.Mumford and J.Shah,“Optimal approximations by piecewise smooth
functions and associated variational problems,”Commun.Pure Appl.
Math.,vol.42,no.5,pp.577–685,Jul.1989.
[17]A.Tsai,A.Yezzi,and A.S.Willsky,“Curve evolution implementation of
the Mumford–Shah functional for image segmentation,denoising,inter-polation,and magni?cation,”IEEE Trans.Image Process.,vol.10,no.8, pp.1169–1186,Aug.2001.
[18]C.Li,C.Kao,J.Gore,and Z.Ding,“Minimization of region-scalable
?tting energy for image segmentation,”IEEE Trans.Image Process., vol.17,no.10,pp.1940–1949,Oct.2008.[19]F.Galland,N.Bertaux,and P.Réfrégier,“Minimum description length
synthetic aperture radar image segmentation,”IEEE Trans.Image Process.,vol.12,no.9,pp.995–1006,Sep.2003.
[20]G.Delyon and P.Réfrégier,“SAR image segmentation by stochastic
complexity minimization with a nonparametric noise model,”IEEE Trans.
Geosci.Remote Sens.,vol.44,no.7,pp.1954–1961,Jul.2006.
[21]C.Fowlkes,S.Belongie,F.Chung,and J.Malik,“Spectral grouping using
the Nystr?m method,”IEEE Trans.Pattern Anal.Mach.Intell.,vol.26, no.2,pp.214–225,Feb.2004.
[22]S.X.Yu and J.Shi,“Multiclass spectral clustering,”in Proc.9th IEEE
https://www.doczj.com/doc/976380975.html,put.Vis.,2003,vol.1,pp.313–319.
[23]C.Ding,X.He,H.Zha,M.Gu,and H.Simon,“A min–max cut algorithm
for graph partitioning and data clustering,”in Proc.1st IEEE ICDM,2001, pp.107–114.
[24]P.R.Kersten,J.-S.Lee,and T.L.Ainsworth,“Unsupervised classi?cation
of polarimetric synthetic aperture radar images using fuzzy clustering and EM clustering,”IEEE Trans.Geosci.Remote Sens.,vol.43,no.3, pp.519–527,Mar.2005.
[25]G.Davidson,K.Ouchi,G.Saito,N.Ishitsuka,K.Mohri,and
S.Uratsuka,“Polarimetric classi?cation using expectation methods,”
in Proc.Polarimetr.Interferometr.SAR Workshop,Tokyo,Japan, Aug.29–30,2002.
[26]J.F.Aujol,G.Aubert,and L.B.Feraud,“Wavelet-based level set evolu-
tion for classi?cation of textured images,”IEEE Trans.Image Process., vol.12,no.12,pp.1634–1641,Dec.
2003.
Gholamreza Akbarizadeh was born in Shiraz,Iran,
on July14,1981.He received the B.S.degree from
the Khajeh-Nassir Tousi University of Technology
(KNTU),Tehran,Iran,in2003and the M.S.and
Ph.D.degrees from the Iran University of Science
and Technology,Tehran,in2005and2011,respec-
tively,all in electrical and electronics engineering.
From2003to2011,he has worked at the DSP
R&D research laboratory as a Senior Researcher.He
is currently an Assistant Professor and Faculty Mem-
ber of the Engineering Department,Shahid Chamran University of Ahvaz,Ahvaz,Iran.He has more than20published papers in different international electrical engineering conferences and journals.His research interests in pattern recognition,machine vision,image processing,and remote sensing analysis.
Dr.Akbarizadeh is member of some international scienti?c societies such as Iranian Machine Vision and Image Processing(MVIP).
带电作业作业步骤及注 意事项 内部编号：（YUUT-TBBY-MMUT-URRUY-UOOY-DBUYI-0128）
(1)人员的准备。作业前应根据作业项目确定操作人员，如作业当天出现某作业人员明显精神和体力不适的情况时，应及时更换人员，不得强行要求作业。(2)工具的准备。作业前应根据作业项目，作业场所的需要，按数配足绝缘遮蔽用具、防护用具、操作工具、运载工具等，并检查是否完好，工器具及防护用具应分别装入规定的工具袋中带往现场。在运输中应严防受潮和碰撞，在作业现场应选择不影响作业的干燥、阴凉位置，分类整理摆放在防潮布上。 绝缘斗臂车在使用前应认真检查其表面状况，若绝缘臂、斗表面存在明显脏污，可采用清洁毛巾或棉纱擦拭，清洁完毕后应在正常工作环境下置放l5min以上，斗臂车在使用前应空斗试操作1次，确认液压传动、回转、升降、伸缩系统工作正常，操作灵活，制动装置可靠。 (3)作业方案的准备。作业负责人应针对作业项目制定作业方案，对较简单的或经常性的作业，可在实施作业的当天在现场制订出作业方案，对较为复杂的作业，应在实施作业的前一天进行调查，研究制订出作业方案，作业负责人应结合方案明确指示每位作业人员的分工，并在班前会议上对作业内容、作业顺序及安全注意事项等作出详细说明。 (4)工器具的检查。到达现场后，在作业前应检查确认在运输、装卸过程中工具有无螺帽松动，绝缘遮蔽用具、防护用具有无破损，应检查、摇测绝缘作业工具。 (5)作业人员在工作现场要仔细检查电杆及电杆拉线，以及上部的腐蚀状况，必要时要采取防止倒塌的措施。应根据地形地貌，将斗臂车定位于最适于作业位置，斗臂车应良好接地，作业人员进入工作斗应系好安全带，要充分注意周边电信线路和高低压线路及其他障碍物，选定绝缘斗的升降回转路径，平稳地操作。
10KV带电作业操作规程 1范围 本标准规定了10kV带电作业的现场操作规程，适用于本标准制定的 10kV带电作业项目。 2人员组织 2.1中间电位法作业（使用绝缘斗臂车），使用一辆绝缘斗臂车时，人员包括工作负责人1名，中间电位电工2名（下分第一、二电工），斗车司机1名，地面电工1名，共5名。使用两辆绝缘斗臂车时，人员包括工作负责人1名，中间电位电工4名（分成两组，每组分第一。二电工），斗车司机2名，地面电工2名，共9名。 2.2地电位法作业，人员包括工作负责人1名，地电位电工2名（下分第一、二电工），地面电工1名，共4名。 3人员的职责范围 3.1 工作负责人 正确安全地组织全组人员工作，结合实际进行安全思想教育，工作前对工作班成员交待工作任务、各成员的职责、安全措施和技术措施。严格执行工作票所列安全措施，必要时还应加以补充。督促、监护工作人员遵守本规程，确认工作班人员变动是否合适。负责核对现场地点、杆号的正确性。在工作的过程中，始终密切监视绝缘斗臂车斗臂升降、转位时的路线；始终密切监视人员登杆路线。 3.2 斗车司机
将斗臂车停放在便于工作的预定位置，确认斗臂液压、传动、回转、升降系统正常，操作灵活，制动装置可靠，并装设临时遮栏。在作业的过程中，应始终保持斗臂车处于发动状态。 3.3 地面电工 应配合斗车司机装设临时遮栏。将所需的安全工器具、材料摆放整齐，检查工器具是否有损伤或损坏并用清洁干燥的毛巾擦拭。 3.4 第一电工 作业时，负责控制车斗的升降、转位和主要的操作。 3.5 第二电工 作业过程中，应对第一电工的工作进行监护，并协助第一电工工作。 3.6 地面工作人员必须共同配合，防止行人进入临时遮栏。 4安全措施 4.1 开展10kV带电作业的条件 4.1.1 中间电位法作业时要求，导线相间距离能满足带电作业的安全距离， 且绝缘斗臂车能够进场作业和带电作业人员具有足够操作空间的10kV架空线路。 4.1.2 地电位法作业时要求， 导线相间距离能满足带电作业的安全距离，带电作业人员具有足够操作空间的10kV架空线路。 4.2 安全注意事项
快速水分测定法的验证 1、概述 药品生产过程中需要水分控制，快速水分测定仪用于水分测定能够缩短水分检验的时间，减少检验人员的劳动强度，降低检验成本，方便管理人员迅速作出决策，从而保证生产工序的持续进行。但快速水分测定仪存在误差，针对这一点特制订本报告，使用快速水分测定仪法与《中国药典2010年版》附录规定的水分测定法进行对比验证。 2、验证目的 对快速水分测定法与《中国药典2010年版》附录规定的水分测定法进行对比验证。通过对比研究确定快速水分测定仪能够有效的保证药品生产过程中对水分的控制，有效地保证药品质量。 3、适用范围 本标准适用于快速水分测定方法的验证。 4、验证领导小组成员及职责 5、验证进度计划 验证小组提出完整的验证计划，经批准后实施。 从年月日至年月日 6、相关文件
2010年版药典一部附录ⅨH 水分测定法；2010年版药典一部附录ⅩⅧA中药质量标准分析方法验证指导原则；实验室控制系统GMP实施指南第11章分析方法的验证和确认；药品生产验证指南第三篇第一章检验方法验证；药品生产质量管理规范（2010年修订）第四章第四节分析方法验证。 7、验证内容 7.1为了确保验证数据的准确可靠，采取以下几个先行保障措施 仪器：已经过校正并在有效期内 人员：均经过培训，熟悉方法及使用的仪器 对照品：均购自中国食品药品鉴定研究院 材料：所用材料，包括试剂、实验用容器等，均符合检验要求，不给实验带来污染、误差。 检查人：日期：确认人：日期： 7.2验证方法 快速水分测定仪设定不同的烘烤温度，不同的烘烤时间。对样品进行水分的测定，并与标准烘箱法作比较，确定最佳烘烤温度，烘烤时间。在此条件下进行方法准确度及精密度的测定。 7.2.1最佳烘烤温度，烘烤时间的确定 选择5批样品分别进行标准烘箱法水分测定以及烘烤温度，烘烤时间下的快速水分测定仪法水分测定。两法相比较，寻求最佳条件。 7.2.1.1最佳烘烤温度的确定 水存在的状态分2种：自由水和结合水。结合水含物理结合水和化学结合水，温度升高，烘干时间减少，误差差异过大，其次有可能造成对化学结合水的破坏，温度过低不适宜于“快速”二字。根据实际生产中对产品水分反应时间的要求，将快速法的烘烤时间定为5min，以6种不同的烘烤温度处理后与标准法比较。实验结果如下： 7.2.1.2 最佳烘烤时间的确定 固定快速法在最佳烘烤温度的条件下，以6种不同的时间处理后与标准法比较。实验结果和分析如下： 品名： *****
油漆喷枪使用说明书 它是我们在喷涂施工中的主要操作设备，根据取料方式以分为三种：虹吸式、重力式和压送式。因除对嘴式喷枪外，不论是其它虹吸式还是重力式都可以很方便的改为压送式，所以我们一般只把喷枪分为两种：虹吸式和重力式。 1.虹吸式喷枪（虹吸式喷枪外部结构如下图）虹吸式喷枪又分为对嘴式和扁嘴式两种。对嘴式喷枪一般重量较轻，仅0.5公斤左右，贮料罐容量小，工作压力一般在０.274～０.345兆帕，喷涂时喷雾呈圆形，喷嘴与工件物面间距150～250毫米，仅适用于小工件的喷涂作业，所以又俗称小喷枪。扁嘴式喷枪相应要重一些，约１公斤，贮料罐容量较大，喷涂气压一般控制在0.35~0.50兆帕，喷涂间距较大，约在250~350毫米，属于大型喷枪。喷涂时喷雾由旋转螺冒调节控制，可呈圆形、扇形或圆锥形（旋紧），喷涂时应根据被涂物表面形状和施工现场情况而调整喷雾形态。 2.压下式喷枪压下式喷枪又称重力式喷枪，其构造和工作原理基本上与扁嘴式喷枪相同，只是贮料罐上方，涂料靠重力作用自行流入枪膛，并随压缩空气喷射在工件上。贮料罐装在上方，即使涂料较少，亦可充分得到利用，这是它的优点。但枪体重心上移，如装满涂料，重量较大，不便于枪体运作，且涂料又易溢出，影响操作。 3.压送式喷枪 压送式喷枪其构造和工作原理基本上与压下式喷枪和虹吸式的扁嘴喷枪一样，其喷枪本身就是由扁嘴喷枪改造而成，不同的只是供
料方式不一样而已，扁嘴式喷枪由贮料罐供料，压送式喷枪采用输漆泵供料。 4.常用喷枪的内部结构及各种控制阀门的作用 ①.枪头扁嘴（旋转螺母）枪头扁嘴主要是用来控制喷涂漆雾的形状和方向。枪头扁嘴与枪身呈垂直或平行时，漆雾方向与枪头扁嘴的方向相反，漆雾呈扇形。枪头扁嘴与枪身呈45°角交差时，漆雾呈圆形，角度变小或增大时，漆雾由圆形向椭圆变化，至180°角时变成扇形。 ②.漆雾面积调节阀漆雾面积调节阀又称扇面调节阀，它在枪体的右上方，流量调节阀的上面。漆雾面积调节阀的主要作用是用来控制漆雾在工件上的施工面积和漆膜的均匀。漆雾面积调节阀向左旋转漆雾面积喷涂面积增大，向右旋转漆雾喷涂面积变小，旋紧时喷涂漆雾呈圆形。 ③. 流量调节阀流量调节阀全名为涂料流量调节阀，以称顶针限动螺杆，它在枪体的右上方，扇面调节阀的下面。流量调节阀是喷涂是用控制涂料喷出的大小，调节喷涂面积和漆膜的厚度、流平。流量调节阀向左旋转出漆量变大，一次性喷涂膜变厚，喷涂面积增大；向右旋转出漆变小，一次性喷涂面积减小，漆膜变薄；旋紧时无涂料喷出。 ④.空气压力调节阀空气压力调节阀以称喷涂空气控制阀，它在枪体的下方，气管连接处的旁边。空气压力调节阀向左旋转，喷涂压力增大，喷涂距离变远，出漆量、喷涂面积同时应该跟着增加，否
水分测定仪（水分测定仪怎么分类）： 能够检测各类有机及无机固体、液体、气体等样品中含水率的的仪器叫做水分测定仪，按测定原理可以分类物理测定法和化学测定法两大类。物理测定法常用的有失重法、蒸馏分层法、气相色谱分析法等，化学测定方法主要有卡尔费休法（Karl Fischer）、甲苯法等，国际标准化组织把卡尔费休（Karl Fischer）方法定为测微量水分国际标准，我们国家也把这个方法定为国家标准测微量水分。 常见的失重法水分仪有卤素水分测定、红外水分测定仪、微波水分测定仪等； 常见的卡尔费休水分测定仪主要有容量法卡尔费休水分测定仪和库仑法（电量法）卡尔费休水分测定仪。 另外还有便携式水份测定仪 红外线水分测定仪： 红外线水分测定仪，采用热解重量原理设计的，是一种新型快速水分检测仪器。水分测定仪在测量样品重量的同时，红外加热单元和水分蒸发通道快速干燥样品，在干燥过程中，水分仪持续测量并即时显示样品丢失的水分含量%，干燥程序完成后，最终测定的水分含量值被锁定显示。与国际烘箱加热法相比，红外加热可以最短时间内达到最大加热功率，在高温下样品快速被干燥，其检测结果与国标烘箱法具有良好的一致性,具有可替代性,且检测效率远远高于烘箱法。一般样品只需几分钟即可完成测定。
仪器操作简单，测试准确，显示部分采用红色数码管，示值清晰可见，分别可显示水分值，样品初值，终值，测定时间，温度初值，最终值等数据，并具有与计算机，打印机连接功能。 水分仪可广泛应用于一切需要快速测定水分的行业，如医药，粮食、饲料、种子，菜籽，脱水蔬菜、烟草，化工，茶叶，食品、肉类以及纺织，农林、造纸、橡胶、塑胶、纺织等行业中的实验室与生产过程中。
500kv紧凑型线路带电作业操作方法 一、更换500kv上相V型合成绝缘子 1.人员组合： 共8人。工作负责人1名、地面电工3名、塔上监护人1名、塔上电工3名。 2.作业方法： 等电位作业与电位作业相结合。 3.操作程序 (1)工作负责人宣讲工作票，并向作业人员交代安全措施后，开始工作。 （2）第一电工携带绝缘无极绳索沿铁塔脚钉攀登至横担上系好安全带，在适当处挂好转相传递滑车，准备传递工作。 （3）塔上电工在地面电工配合下，在横担吊线处安装横担固定器、提线器、绝缘吊杆、二道保护的双钩、杆、提线卡具等。 （4）第一、第二电工将传递上来的绝缘硬梯挂在横担下平面中间通道上，绝缘梯中不应用绝缘绳固定在塔身上，以防吊梯摆动。 （5）在检查绝缘梯悬挂无误，第三得到塔上监护人许可后系好人身二道保护绳沿绝缘梯下到绝缘梯地部（即上导线水平位置），保护绳由第二电工控制，保护绳随走随放，监护人时刻注意安全情况。 （6）第三电工下到绝缘梯底部后，系好安全带，在一次检查梯拉绳和吊梯安全无误后，将随身携带下的梯拉绳和吊梯可靠连接做好进入电场准备。 （7）得到工作负责人下达进入等电位命令后，塔上第一、第二电工听从塔上监护人统一指挥，解开控制吊梯的防晃绳，控制吊梯拉绳方向，将吊绳拉向需工作导线侧，进入等电位工作状态。系好安全带后方可工作，在工作过程中，头部不得超越均压环，保持安全距离。（8）第三电工将绝缘吊杆连板卡具，提线卡具与绝缘吊杆连接好。 （9）二道保护绝缘吊杆与提线卡具连接到分裂导线两根导线上，合成绝缘子高强度绝缘保护绳固定在合成绝缘子串连接点下侧以防合成绝缘子松开后摆动。 （10）提升导线时两提升器速度要一致，时刻注意各部位情况，按第三电工要求将导线提升到位。 （11）收紧二道保护，收紧防摆动高强度牵引绝缘绳，拆除需要更换的绝缘子串。 （12）塔上电工利用无极绳将合成绝缘子垂直传向地面进行更换。 （13）恢复绝缘子串与上述程序相反方向进行，在检查无误后拆除工具，退出带电体，塔上人员下到地面，工作结束。 4.主要专用工具 绝缘梯（6m）一副、绝缘吊线杆（5m）二根、提线器及横担固定器二副、提线卡具一副、连板卡具一副、高强度芳纶保护绳一根、电动提升机一台、二道保护双钩一副、保护绝缘吊杆（4.2m）一根、绝缘控制绳一根。 5.安全注意事项： （1）所有工具在使用前必须检查良好方可使用。 （2）在导线二道保护用提线卡具未承受导线荷载前，不得拆开被换合成绝缘子。
2． 1."4仪器预热方式，在使用前必须预热30分钟，若环境温度较低需预热一小时，经预热后测定的数据真实有效。 第三章快速水分测定仪操作规程 3.1取样与样品制备 3. 1."1取样与样品的制备，与重要测试参数的选择对最终的测试精度十分关键。 3. 1."2凡被测样品颗粒状的，应用粉碎机粉碎成粉状，方可进行测定，否则测试时间过长，且水分含量不能完全被干燥。 3. 1."3在采集颗粒，粉状样品过程中，一定要充分混样，保证样品水分均匀。 3. 1."4对于一些易燃，易爆的测试样品，应选择较低的加热温度，并尽量取少的样品量，对于这些样品的测试应十分小心。 3. 1."5对于大水分含量（粮食或液体）的样品制备过程中，应尽量避免水分丧失，若不能避免，这部分丧失应计算进去。 3. 1."6对于柔软，粘弹性的样品，应切割成尽量薄的片。 3.
1."7取样量的多少，建议用户一个基本原则就是薄薄铺满称盘底面积的量，粉状样品大致为3克左右。 3. 1."8每次取样量的精度也很重要，误差应小于± 0."005xx。 3.2测定水分操作 3. 2."1在仪器现时重量指示灯亮状态下，用手扶住机身，轻轻掀起加热筒。用试样匙取试样放在称量盘中，此时仪器显示“ 3.000”克左右（取样数量对测定精度有一定影响，其规律是取样量大，重复性好，但测定时间长，兼顾精度与时间，我们建议取样为3± 0."005克），取下称量盘用手将试样表面抖均匀或用小棒使试样表面均匀（注意防止试样丢失），放在托架上，连同托架一起轻轻放到称量盘支架上，合上加热筒。 3. 2."2待重量显示稳定时，按“↑”键，紧接按“测试”键，仪器自动开始测定水分，此时水分示值指示灯亮，数据窗显示正在失去的水分量。温度显示值在上升，直到设定值。 在测定水分中温度显示值上下跳动2-3度为正常现象。 3. 2."3当水分测定完成后，仪器自动停止加热，并发出报警声，按一次“显示”键即可消除报警声，此时显示判别时间。再按一次“显示”键，仪器显示最终水分值。若仪器连接打印机则直接按“打印”键打印出水分值。需要查看其它测试参数可连续按“显示”依次查看，最后按“清除”键使仪器回到现时重量状态下。
带电作业操作导则 1 总则 2 带电作业的有关术语 3 带电作业操作的一般要求 4 地电位作业的基本要求 5 等电位作业的基本要求 6 中间电位作业的基本要求 7 常规作业项目的操作 1 总则 1.1为确保带电作业安全，实现带电作业规范化、标准化，在总结我国开展带电作业四十多年经验的基础上，制订了本《带电作业操作导则》(以下简称《导则》)。 1.2本《导则》适用于交流10～500kv输配电线路和变电所(开闭所、发电厂)电气设备上的带电作业项目。 1.3本《导则》提出了项目的操作方法、工(器)具使用范围、操作要领和安全注意事项，同时，对带电作业方法、项目和工(器)具有关术语作了统一规定。 1.4鉴于我国电气设备型式多样，作业项目开发较多，从安全和实用出发，本《导则》正文只推荐常用项目，且操作方法只作原则指导。 1.5各基层单位(指电业局、供电局、供电公司、发电厂、发电公司等，下同〉应按本《导则》的规定，根据各自的实际情况编制出适合本单位的《带电作业现场操作规程》。 1.6《带电作业现场操作规程》应按项目制订。其内容应包括: a.项目名称； b.适用范围； c.作业方法； d.劳动组合； e.操作步骤； f.安全措施； g.所需工具。 较为复杂的项目还应附有操作示意图。 1.7 dl409、dl408～91《电业安全工作规程》(以下简称《安规》)中的带电作业章节和其它有关专业标准中已提及的安全措施，本《导则》一般不再重复，因此，各基层单位除应遵守本《导则》规定外，还应严格遵守《安规》和其它有关专业标准中的条文规定。 2 带电作业的有关术语
2.1作业方法 根据作业时作业人员所处的电位高低，带电作业方法可分为地电位作业法等电位作业法和中间电位作业法三种，如图1所示。 图1 带电作业方法分类示意图 (a)地电位作业法；(b)等电位作业法；(c) 中间电位作业法 1- 带电体；2-绝缘体；3-人体 2.1.1地电位作业法：是指作业人员于地电位，通过绝缘工具对带电体进行的 作业。该法又称"零电位作业法"。 2.1.2等电位作业法：是指作业人员通过绝缘体对大地绝缘后，人体与带电体处于同一电位下进行的作业。 沿绝缘子串进入强电场或在绝缘子串上作业，是等电位作业的一种特殊方式。 2.1.3中间电业作业法:是指作业人员通过绝缘体对大地绝缘而同时又与带电体保持一定距离情况下，用较短的绝缘工具对带电体进行的作业。 2.2作业项目 作业项目是指利用某种操作方法，在设备不停电的情况下完成某项具体工作。 项目的名称应包括以下内容: a.作业设备名称；
SFY-20红外线快速水分测定仪 使用说明书 第一章概述 首先感谢您选用本公司生产的SFY-20红外线快速水分测定仪。请您在使用前详细阅读本说明书，如有疑问，可向经销商咨询或和本公司联系。 1.1用途、特点 SFY-20红外线快速水分测定仪，采用热解重量原理设计的，是一种新型快速水分检测仪器。水分测定仪在测量样品重量的同时，红外加热单元和水分蒸发通道快速干燥样品，在干燥过程中，水分仪持续测量并即时显示样品丢失的水分含量%，干燥程序完成后，最终测定的水分含量值被锁定显示。与国际烘箱加热法相比，红外加热可以最短时间内达到最大加热功率，在高温下样品快速被干燥，其检测结果与国标烘箱法具有良好的一致性，具有可替代性，且检测效率远远高于烘箱法。一般样品只需几分钟即可完成测定。该仪器操作简单，测试准确，显示部分采用红色数码管，示值清晰可见，分别可显示水分值，样品初值，终值，测定时间，温度初值，最终值等数据，并具有与计算机，打印机连接功能。因此该水分仪可广泛应用于一切需要快速测定水分的行业，如医药，粮食、种子，菜籽，烟草，化工，茶叶，食品、肉类、种子、石墨、油墨、锯末、沙土、砂石以及纺织，农林、造纸、橡胶、塑胶等行业中的实验室与生产过程中。 1.2 SFY-20主要技术指标 水分测定范围（%）： 0.01%-100% 测定试样重量（g）： 0-90 最大称重量：（g）： 20 称量最小读数（g）: 0.001 水分含量可读性(％): 0.01 温度设定范围（℃）：室温-160 显示参数： 7种 通讯接口：标准RS232接口 波特率：9600/S比特 通讯方式：MCS51系列单片机通讯方式2。 供电电源：电压220v±10%频率50HZ±1HZ 试样温度：-40℃-50℃ 工作环境温度：-5℃-50℃ 相对湿度：≤80%RΗ 外形尺寸：380mm×205mm×325mm 净重量：3.7kg
油漆喷枪使用说明书它是我们在喷涂施工中的主要操作设备，根据取料方式以分为三种：虹吸式、重力式和压送式。因除对嘴式喷枪外，不论是其它虹吸式还是重力式都可以很方便的改为压送式，所以我们一般只把喷枪分为两种：虹吸式和重力式。 1.虹吸式喷枪（虹吸式喷枪外部结构如下图）虹吸式喷枪又分为对嘴式和扁嘴式两种。对嘴式喷枪一般重量较轻，仅公斤左右，贮料罐容量小，工作压力一般在０.274～０.345兆帕，喷涂时喷雾呈圆形，喷嘴与工件物面间距150～250毫米，仅适用于小工件的喷涂作业，所以又俗称小喷枪。扁嘴式喷枪相应要重一些，约１公斤，贮料罐容量较大，喷涂气压一般控制在~兆帕，喷涂间距较大，约在250~350毫米，属于大型喷枪。喷涂时喷雾由旋转螺冒调节控制，可呈圆形、扇形或圆锥形（旋紧），喷涂时应根据被涂物表面形状和施工现场情况而调整喷雾形态。 2.压下式喷枪压下式喷枪又称重力式喷枪，其构造和工作原理基本上与扁嘴式喷枪相同，只是贮料罐上方，涂料靠重力作用自行流入枪膛，并随压缩空气喷射在工件上。贮料罐装在上方，即使涂料较少，亦可充分得到利用，这是它的优点。但枪体重心上移，如装满涂料，重量较大，不便于枪体运作，且涂料又易溢出，影响操作。 3.压送式喷枪
压送式喷枪其构造和工作原理基本上与压下式喷枪和虹吸式的扁嘴喷枪一样，其喷枪本身就是由扁嘴喷枪改造而成，不同的只是供料方式不一样而已，扁嘴式喷枪由贮料罐供料，压送式喷枪采用输漆泵供料。? 4.常用喷枪的内部结构及各种控制阀门的作用 ①.枪头扁嘴（旋转螺母）枪头扁嘴主要是用来控制喷涂漆雾的形状和方向。枪头扁嘴与枪身呈垂直或平行时，漆雾方向与枪头扁嘴的方向相反，漆雾呈扇形。枪头扁嘴与枪身呈45°角交差时，漆雾呈圆形，角度变小或增大时，漆雾由圆形向椭圆变化，至180°角时变成扇形。 ②.漆雾面积调节阀漆雾面积调节阀又称扇面调节阀，它在枪体的右上方，流量调节阀的上面。漆雾面积调节阀的主要作用是用来控制漆雾在工件上的施工面积和漆膜的均匀。漆雾面积调节阀向左旋转漆雾面积喷涂面积增大，向右旋转漆雾喷涂面积变小，旋紧时喷涂漆雾呈圆形。 ③. 流量调节阀流量调节阀全名为涂料流量调节阀，以称顶针限动螺杆，它在枪体的右上方，扇面调节阀的下面。流量调节阀是喷涂是用控制涂料喷出的大小，调节喷涂面积和漆膜的厚度、流平。流量调节阀向左旋转出漆量变大，一次性喷涂膜变厚，喷涂面积增大；向右旋转出漆变小，一次性喷涂面积减小，漆膜变薄；旋紧时无涂料喷出。
第一节直线绝缘子串 (2) 1.1 110kV输电线路带电更换悬垂绝缘子串 (2) 1.2 110kV输电线路带电更换悬垂绝缘子串 (4) 1.3 110kV输电线路带电更换直线悬垂绝缘子串 (6) 1.4 110kV输电线路带电清扫绝缘子 (8) 第二节耐张绝缘子串 (9) 2.1 110kV输电线路带电更换耐张整串绝缘子 (10) 2.2 110kV输电线路带电更换双串耐张整串绝缘子 (11) 2.3 110kV输电线路带电更换耐张整串绝缘子 (13) 2.4 110kV输电线路带电更换跳线（引流线）绝缘子串 (15) 第三节金具及附件 (17) 3.1 110kV输电线路带电更换导线悬垂线夹 (17) 3.2 110kV输电线路带电更换耐张跳线（引流线）并沟线夹 (19) 3.3 110kV输电线路带电更换导线防振锤（防舞动失谐摆） (21) 3.4 110kV输电线路带电更换子导线间隔棒（环） (22) 3.5 110kV输电线路带电安装（更换）导线相间间隔棒 (24) 3.6 110kV输电线路带电安装导线防舞鞭 (25) 3.7 110kV输电线路带电安装（更换）双摆防舞器 (27) 3.8 110kV输电线路安装防鸟剌、防鸟网、消雷器（避雷针） (28) 3.9 110kV输电线路带电安装防鸟罩 (30) 第四节导、地线 (31) 4.1 110kV输电线路带电修补导线 (31) 4.2 110kV输电线路带电修补导线 (33) 4.3 110kV输电线路带电处理导线异物 (35) 4.4 110kV输电线路带电更换孤立档架空地线 (36) 4.5 110kV输电线路带电处理架空地线轻微损伤 (38) 4.6 110kV输电线路带电进行架空地线断股补强 (39) 第五节检测 (41) 5.1 110KV输电线路带电检测零值瓷绝缘子 (41) 5.2 110kV输电线路复合绝缘子憎水性检测 (42) 5.3 110KV输电线路带电检测低、零值瓷质绝缘子 (44) 5.4 带电检测110 kV线路绝缘子等值附盐密度、灰密 (46) 第6节杆塔换部件 (48) 6.1 110KV输电线路带电更换杆塔拉线 (48) 6.2 110KV输电线路带电更换直线∏型杆中、下段 (50) 6.3 110KV输电线路带电升高Π型直线杆导线横担及加装地线支架 (52) 6.4110kV输电线路带电升高或下沉直线Π型电杆 (54) 6.5 110KV输电线路带电升高或下沉直线单杆 (55) 第7节特殊或大型作业项目 (57) 7.1 110KV输电线路带电断开和接通空载线路 (57) 7.2 110KV输电线路带电加装耐张引流线吊串绝缘子 (59) 7.3 110KV输电线路带电调整弛度 (61) 7.4 110KV输电线路跨越110kV带电线路导线展放 (63) 7.5 110KV输电线路带电加装重锤（悬垂绝缘子串、耐张吊瓶） (65) 7.6 110KV输电线路绝缘子带电水冲洗 (66) 7.7 110KV输电线路带电顺线路整体移动门型混凝土杆 (68) 7.8 输电线路档中带电开断损伤导线重新压接作业法 (70)
操作规程编号：YTO-FS-PD726 水分快速测定仪操作规程通用版 In Order T o Standardize The Management Of Daily Behavior, The Activities And T asks Are Controlled By The Determined Terms, So As T o Achieve The Effect Of Safe Production And Reduce Hidden Dangers. 标准/ 权威/ 规范/ 实用 Authoritative And Practical Standards
精品规程范本 编号：YTO-FS-PD726 2 / 2 水分快速测定仪操作规程通用版 使用提示：本操作规程文件可用于工作中为规范日常行为与作业运行过程的管理，通过对确定的条款对活动和任务实施控制,使活动和任务在受控状态，从而达到安全生产和减少隐患的效果。文件下载后可定制修改，请根据实际需要进行调整和使用。 1、水分快速测定应符合《选煤厂技术检查规定》的要求，适用于生产过程中产品水分的快速测定。 2、工作前，先对水分快速测定仪进行预热。 3、用预先干燥过的测定盘迅速称取粒度小于6mm 的煤样10—12g （称准到0.01g ），平摊在测定盘中。 3、打开测定仪，放入煤样，按下测定键，测定仪自动工作。 4、测定过程中不得打开测定仪或中断作业，否则本次测定无效。 5、待测定结束，自动显示水分和测定时间。 6、及时记录测定结果。 7、工作结束，进行断电，盖好防尘罩。 该位置可输入公司/组织对应的名字地址 The Name Of The Organization Can Be Entered In This Location
带电作业现场操作规程 （试行） 目录 一总则 (2) 二、组织措施 (2) 三、技术措施 (3) 四、常用项目操作规程 (5) 4-1带电断、接空载线路引流线（地电位作业） (5) 4-2带电断、接空载线路引流线（等电位作业） (6) 4-3带电更换刀闸、高压熔断器 (7) 4-4带电更换直线杆中相针式绝缘子 (8) 4-5带电更换直线杆边相针式绝缘子 (9) 4-6带电更换单回直线横担 (10) 4-7带电更换耐张绝缘子 (11) 4-8设备连接点发热带电加分流线 (13) 4-9带电修补导线 (13) 4-10带电去除导线上导物 (14) 五、高空绝缘斗臂车上导物 (15) 附录：带电作业用高空绝缘斗臂车库房标准 (16)
一总则 1．1 为确保配电线路带电作业安全，逐步实现配网带电作业规范化、标准话、根据国网公司《电业安全工作规程》带电作业部分，国家电力公司《带电作业操作导则》，总结先进地区电力公司带电作业工作经验，特制定了《右中电力公司10KV配网带电作业现场操作规程（试行）》（以下简称《规程》）。 1．2 本规程适用于右中电网内10KV配网带电作业的所有工作人员。 1．3 所有参加10KV配网带电作业的人员、专业管理人员及有关领导必须熟悉并严格遵守本规程、国网公司颁布的《电业安全工作规程》带电作业部分。1．4 本规程只包括常用项目，对于新项目和新工具应经电气性能、机械强度试验合格，制定出安全措施和操作工艺方案并在模拟设备上试验无误后，经公司主管生产副经理（总工程师）批准方可实行和使用。 1．5 带电作业人员应从电力技校、中专毕业生或从线路施工、运行、检修技术熟练的年青技工中择优挑选，经过带电作业知识和实际模拟操作技术培训、考试合格、经人事劳动部门批准取得带电作业合格证后，方能参加带电作业工作。1．6 带电作业人员应保持相对稳定，未经公司主管生产副经理（总工程师）批准，不得任意调动。 二组织措施 2．1 有权签发带电作业工作票人员名单，应经安监部门审核后报公司领导批准，并颁布。工作票签发人应熟悉系统，熟悉人员技术水平，并对以下事项负责：2．1．1所进行的工作是否符合安全规程； 2．1．2工作票上填写的安全技术措施是否正确完善； 2，1．3所派工作负责人和工作人员是否适当和充足。 2．2带电作业工作负责人职责； 2．2．1制定带电作业安全措施；
SFY-20A卤素快速水分测定仪 使用说明书 第一章概述 首先感谢您选用本公司生产的SFY-20A卤素快速水分测定仪。请您在使用前详细阅读本说明书，如有疑问，可向经销商咨询或和本公司联系。 1.1用途、特点 SFY-20A卤素快速水分测定仪，是一种新型快速水分检验仪器，可用来测量任何物质的水分含量，（通过加热发生危险化学反应的物质除外）该仪器采用热解重量原理设计的，仪器测量样品重量同时，卤素加热单元和水分蒸发通道快速干燥样品。在干燥过程中，仪器持续测量并即时显示干燥过程中样品丢失的水分含量％，干燥完成后，最终测定的水分含量锁定，按显示键可观察水分值，重量初始值，起始值，测试时间等数据。与传统的烘箱加热法相比，卤素加热可在最短时间达到最大加热功率，在高温下样品快速被干燥，大大加快了测量时间。该仪器可用于一切需要快速测量水分的行业，如医药，粮食，烟草，化工，茶叶，食品，纺织，农林等。该仪器可与计算机通讯，并通过计算机把测试水分数据结果打印出来。也可通过选配的打印机直接将测试水分数据结果打印出来。 1.2 SFY-20A主要技术指标 水分测定范围（%）： 0.01%-100 测定试样重量（g）： 0-90 最大称重量：（g）： 20 称量最小读数（g）: 0.001 水分含量可读性(％): 0.01 温度设定范围（℃）：室温-160 显示参数： 7种 通讯接口：标准RS232接口 波特率：9600/S比特 通讯方式：MCS51系列单片机通讯方式2。 供电电源：电压220v±10%频率50HZ±1HZ 试样温度：-40℃-50℃ 工作环境温度：-5℃-50℃ 相对湿度：≤80%RΗ 外形尺寸：337mm×215mm×334mm 净重量：3.7kg 第二章调试 2.1测定前准备工作 2.1.1仪器放在固定且水平的工作台上，把仪器底面保护称重部件锣丝拧松至锣丝头部抵到桌面，掀开加热筒，轻轻放上称量支架与称量盘，调节仪器后底板下二个活动脚，使仪器水平。 2.1.2检查交流电源插座完好，取出电源线，与仪器接通电源。
带电作业操作的一般要求及安全注意事项 摘要：带电作业是指在没有停电的设备或线路上进行的工作。实施带电作业，在不间断供电的前提下处理电网故障与缺陷，有助于迅速完成检修工作，提高供电可靠性。本文介绍了带电作业的一般要求以及安全注意事项，对输电专业管理进行了系统分析，并且结合生产单位的实际情况，提出带电作业操作的一般要求及安全注意事项，详细对专业性质和人员素质提出了新的方案。 1 概述 输电专业是一项极为普通而又复杂的专业，是电力行业的基础，也是电力行业的排头兵。而带电作业是输电专业一项特殊的工种，是在不停电的前提下对输电设备进行检修作业，是输电设备能够安全稳定的有力保障。带电作业可以提高供电可靠性，增多供电量，从而增收了经济效益，为增供扩销工作打下了坚实的基础。 随着市场经济的快速发展，人们对供电可靠性要求也越来越高，用电量也在与日俱增。如果因设备缺陷而实行停电检修，短时间的停电也会给用户的生产生活造成较大的负面影响，实施带电作业，在不间断供电的前提下处理电网故障和缺陷，可有效解决与用户用电之间的矛盾，有助于迅速地完成检修工作，提高供电可靠性。因此，全面持久的开展带电作业对提高电网的安全运行至关重要，下面我就带电作业操作的一般要求及安全注意事项进行分析。 2带电作业准备工作 带电作业指的是在没有停电的设备或线路上进行工作，如在带电的电气设备上或线路上用特殊的方法进行测试、维护、检修和个别零部件的拆装工作。按照带电作业人员按作业是否直接接触带电体，可以分为直接作业和间接作业。按作业人员作业时所处的电位高低，可以分为等电位作业、中间电位作业以及地电位作业。目前，带电作业采取的主要方式有间接带电作业、等电位作业、沿绝缘子串进入强电场作业、分相作业和全电缆作业等。带电作业的准备工作主要包括以下几个方面： 1) 带电作业班组在接受带电作业后，应根据任务难易和对作业设备熟悉程度，决定是否需要查阅资料和查勘现场。 (1) 查阅资料：是指从生产部、运行班组和资料室了解作业设备的情况。如导、地线规格、设计所取的安全系数及荷载；杆塔结构、档距；系统结线、相位和运行方式；设备状况（指导、地线补强、锈蚀、接头等）及作业环境情况，以便根据作业内容确定作业方法、所需工（器）具，并作出是否需要停用重合闸的决定。必要时还应作如下工作：首先验算导、地线应力，或计算导地线张力或悬垂重量；其次计算空载电流、环流和电位差；最后计算悬重后的弧垂，并校核对地或被跨越物的安全距离。 (2) 现场勘查：赴作业现场主要了解作业设备各种间距、交叉跨越、缺陷部位以及严重程度、地形状况、周围环境、确定需用器材及工（器）具等。根据现场勘查结果，作出能否进行带电作业、采用何种作业方法及必要的安全措施等决定。 2) 带电作业班组去现场前，应注意当地气象部门的当天天气预报。到达现
国标法烘干法快速水分测定仪说明书 概述 水是万物之源，绝大多数物质里面都有水的存在。在生产加工中，水分含量是大多数行业必须检测的指标之一，目前水分检测在现行的国家标准里面，干燥失重法适用于大多数的行业，本文简单介绍了SFY-20D国标烘干法快速水分测定仪，供读者参考， 水分测定仪工作原理 采用干燥失重法原理，通过加热系统快速加热样品，使样品的水分能够在最短时间之内完全蒸发，从而能在很短的时间内检测出样品的含水率。检测一般样品通常只需3分钟左右。冠亚水分仪采用的原理与国家标准烘箱法相同，检测结果具有可替代性，仪器采用一键式操作，不仅操作简单而且也避免了人为因素对测量结果产生的误差。 水分测定仪操作步骤 第一步：按校准键，放砝码，自动校准。（定期效准，不用每天开机效准） 第二步：取样xg，按测试键开始工作。 第三步：仪器加热中，仪器正在显示丢失的水分值。 第四步：测定结束，仪器显示最终水分。 水分测定仪特点 检测速度快，只需几分钟，创行业之最； 采用最新一代传感技术,快速、简便,一键式操作; 操作简单，全自动操作模式，无可动部件； 关键零部件均采用纯进口高端材料，以保证产品检测结果的准确性； 零易损件，样品盘采用耐酸耐碱耐变形的纯不锈钢材料，无易耗品，样品盘克循环利用； 采用特质的环形卤素光源，加热均匀，加热器更耐用；
水分测定仪技术参数 1、称重范围：0-90g 可调试测试空间为3cm 2、水分测定范围：0.01-100% 3、样品质量：0.100-90g 4、加热温度范围：起始-205℃ 加热方式：可变混合式加热 微调自动补偿温度最高15℃ 5、水分含量可读性：0.01% 6、显示参数：7种 红色数码管独立显示模式 7、外型尺寸：380×205×325(mm) 8、电源：220V±10% 9、频率：50Hz±1Hz 10、净重：3.7Kg 水分测定仪使用注意事项 1.在测定水分过程中,一定要避免震动,加热筒下端缺口不能迎风摆放。 2.测定样品在称量盘中堆积一定要平整,堆积面积尽量布满称盘底面,堆积厚度应尽量薄,利于水分完全蒸发。 3.在测定水分过程中,不能用手去摸加热筒,严禁敲击或直接振动工作台面。 4.由于该仪器称重系统为精密设备,尤其传力部分特别怕重压,冲击,因而在每次取,放称量盘时尽量用托架,若用手进行取,放称量盘应轻取,轻放。 5.测定完成后,马上取下称量盘必须用托架,以免烫手.托架在放入仪器中不应碰到称重支架与称量盘。 6.测定后须待称量盘完全冷却后,再放入下一个试样。
HECY-SE 系列高效脱硝喷枪 安装前请仔细阅读 河北诚誉喷雾技术有限公司安装使用说明书
目录 1喷枪简介 ........................................... - 0 -结构及特点....................................... - 0 -适用范围......................................... - 1 -2喷枪安装 ........................................... - 1 -套筒的加工........................................ - 1 -确定位置开孔...................................... - 1 -固定套筒.......................................... - 2 -安装喷枪......................................... - 2 -喷枪连接.......................................... - 5 -主气路连接........................................ - 6 -3注意事项 ........................................... - 6 -4常见问题及处理方法.................................. - 7 -5服务指南 ........................................... - 7 -
带电作业现场操作 规程
带电作业现场操作规程 （试行） 目录 一总则 (2) 二、组织措施 (2) 三、技术措施 (3) 四、常见项目操作规程 (5) 4-1带电断、接空载线路引流线（地电位作业） (5) 4-2带电断、接空载线路引流线（等电位作业） (6) 4-3带电更换刀闸、高压熔断器 (7) 4-4带电更换直线杆中相针式绝缘子...................................8 4-5带电更换直线杆边相针式绝缘子.. (9) 4-6带电更换单回直线横担 (10) 4-7带电更换耐张绝缘子 (11) 4-8设备连接点发热带电加分流线....................................13 4-9带电修补导线 . (13) 4-10带电去除导线上导物 (14)
五、高空绝缘斗臂车上导物..........................................15附录：带电作业用高空绝缘斗臂车库房标准 (16) 一总则1．1 为确保配电线路带电作业安全，逐步实现配网带电作业规范化、标准话、根据国网公司《电业安全工作规程》带电作业部分，国家电力公司《带电作业操作导则》，总结先进地区电力公司带电作业工作经验，特制定了《右中电力公司10KV配网带电作业现场操作规程（试行）》（以下简称《规程》）。1．2 本规程适用于右中电网内10KV配网带电作业的所有工作人员。1．3 所有参加10KV配网带电作业的人员、专业管理人员及有关领导必须熟悉并严格遵守本规程、国网公司颁布的《电业安全工作规程》带电作业部分。1．4 本规程只包括常见项目，对于新项目和新工具应经电气性能、机械强度试验合格，制定出安全措施和操作工艺方案并在模拟设备上试验无误后，经公司主管生产副经理（总工程师）批准方可实行和使用。1．5 带电作业人员应从电力技校、中专毕业生或从线路施工、运行、检修技术熟练的年青技工中择优挑选，经过带电作业知识和实际模拟操作技术培训、考试合格、经人事劳动部门批准取得带
10KV带电作业现场操作规程 1 范围 本标准规定了10kV带电作业的现场操作规程，适用于本标准制定的10kV带电作业项目。 2 人员组织 2.1中间电位法作业（使用绝缘斗臂车），使用一辆绝缘斗臂车时，人员包括工作负责人1名，中间电位电工2名（下分第一、二电工），斗车司机1名，地面电工1名，共5名。使用两辆绝缘斗臂车时，人员包括工作负责人1名，中间电位电工4名（分成两组，每组分第一。二电工），斗车司机2名，地面电工2名，共9名。 2.2地电位法作业，人员包括工作负责人1名，地电位电工2名（下分第一、二电工），地面电工1名，共4名。 3 人员的职责范围 3.1 工作负责人 正确安全地组织全组人员工作，结合实际进行安全思想教育，工作前对工作班成员交待工作任务、各成员的职责、安全措施和技术措施。严格执行工作票所列安全措施，必要时还应加以补充。督促、监护工作人员遵守本规程，确认工作班人员变动是否合适。负责核对现场地点、杆号的正确性。在工作的过程中，始终密切监视绝缘斗臂车斗臂升降、转位时的路线；始终密切监视人员登杆路线。 3.2 斗车司机 将斗臂车停放在便于工作的预定位置，确认斗臂液压、传动、回转、升降系统正常，操作灵活，制动装置可靠，并装设临时遮栏。在作业的过程中，应始终保持斗臂车处于发动状态。 3.3 地面电工 应配合斗车司机装设临时遮栏。将所需的安全工器具、材料摆放整齐，检查工器具是否有损伤或损坏并用清洁干燥的毛巾擦拭。 3.4 第一电工 作业时，负责控制车斗的升降、转位和主要的操作。 3.5 第二电工 作业过程中，应对第一电工的工作进行监护，并协助第一电工工作。 3.6 地面工作人员必须共同配合，防止行人进入临时遮栏。