基于图像特征的铜粗选过程病态工况识别
2014-09-18卢明桂卫华彭涛谢永芳
卢明+桂卫华+彭涛+谢永芳
收稿日期:20130917
基金项目:国家创新研究群体科学基金资助项目(61321003);国家自然科学基金重点资助项目(61134006);国家自然科学基金资助项目(61273169)
作者简介:卢明(1979-),男,湖南益阳人,中南大学博士研究生
通讯联系人,Email: mlu@hnust.edu.cn
摘要:泡沫图像特征是指泡沫图像中与浮选性能相关的局部黑色水化区域大小,即局部光谱特征.针对这一局部光谱特征形状、大小无规则性,提出了一种基于多维主元分析的特征提取方法,并将提取的特征应用于铜浮选粗选过程病态工况识别.首先,描述了铜浮选粗选过程,分析了影响粗选过程的主要因素和黑色水化区域形成机理;然后,提出一种基于多维主元分析的图像局部光谱特征提取方法;最后,将基于多维主元分析的图像局部光谱特征提取算法应用于铜浮选粗选泡沫图像,并将所提取的图像特征用于铜粗选病态工况识别.工业现场数据验证了所提方法的有效性.
关键词:泡沫图像;图像特征;多维主元分析(MPCA);病态工况识别;铜粗选过程
中图分类号:TP391.41 文献标识码:A
Sick Condition Recognition Based on the Image
Feature of Froth Image in Copper Rough Process
LU Ming1,2,GUI Weihua1,PENG Tao1,XIE Yongfang1
(1.School of Information Science and Engineering, Central South Univ, Changsha, Hunan410083,China;
2.School of Information and Electrical Engineering, Hunan Univ of Science and Technology, Xiangtan, Hunan411201,China)
Abstract:The image features of copper flotation froth image means the size of the area of local black hydration in the froth image, which is called local spectral feature and related to flotation performance. A local spectral feature extraction method based on MPCA was proposed for the irregularity of the size and the shape, and the extracted features were used in copper rougher flotation process to identify sick conditions. Firstly, we described the copper rougher flotation process and analyzed the impact of the main factors roughing process and the formation mechanism of black hydration region. Then, a method was proposed to extract the local feature of image based on MPCA. Lastly, the image local feature extraction algorithm based on MPCA was applied to the copper flotation rougher froth image and the extracted image features were used in copper rougher process for sick condition recognition. The validity of the proposed method has been verified with industrial data.
Key words:froth images;image feature;MultiPrincipal Component(MPCA);sick condition recognition;copper rough process
浮选是一种应用最为广泛的将有用矿物从矿石中分离出来的选矿方法.一直以来, 选厂的生产操作都是依靠有经验的工人对浮选泡沫进行肉眼观察完成的,对泡沫的判断缺乏客观标准, 使得人工观测为主的矿物浮选过程难以处于稳定最优运行状态[1-2].采用机器视觉代替人类视觉, 利用图像处理技术从泡沫图像中提取出最为显著、有效的视觉特征,对浮选泡沫进行客观描述, 并将视觉特征应用于浮选过程的工况识别,能为矿物浮选过程实现实时控制与优化提供操作指导[3-5].
浮选流程大多分为粗选、扫选、精选3个流程单元,每个流程单元由数量不等的浮选槽组成,各个流程之间彼此连接,相互影响[6-8].其中粗选首槽浮选工况好坏,直接影响了后续流程的操作和最终的产品质量及产能.在整个流程中,粗选过程工况的识别尤为重要.以粗选首槽泡沫品位为评价指标,将铜粗选工况分为“正常”和“病态”两个区域.铜粗选过程中的“病态”工况是指因初始条件和操作条件改变而导致粗选产品质量不能满足后续浮选流程要求的工况.当出现病态工况时,浮选泡沫图像中的泡沫颜色(光谱)和形态特征会发生相应的变化.
本文描述了铜浮选粗选过程的特点,提出以黑色水化区域面积作为铜浮选粗选泡沫图像局部光谱特征,并针对这一特征大小、形状无规则性,提出一种基于MPCA的局部光谱特征提取新方法,并将所提取的特征用于铜浮选粗选病态工况识别.
1铜浮选粗选过程描述及泡沫图像局部光
谱特征
如图1所示,为某铜浮选厂生产流程,分为粗选、扫选、精选3个流程单元.虚线框为粗选过程.在整个铜浮选流程中,粗选是矿石经过磨矿、注水、分级后进入选别的第1步.粗选首槽的浮选工况好坏,直接影响了后续流程的操作和最终的产品质量及产能.粗选工况好坏的衡量指标是粗精矿品位,根据冶金学工业试验,粗选的泡沫品位不能太高也不能太低,需要控制在某一范围内,超出这一范围, 则视为粗选过程工况处于“病态”,需要及时调整操作变量.长期以来,粗选过程的操作依赖于“人工看泡”[9-11].但是浮选现场环境恶劣,劳动强度大,而且人工判别的方式主观性太强,易导致工况波动.如图1所示,在粗选首槽安装CCD彩色摄像机获取粗选首槽泡沫图像,从泡沫图像中提取出最为显著、有效的视觉特征,并将所提取的特征用于铜粗选过程病态工况识别,可以规范操作,为后续流程的调整提供指导.
铜粗选过程中的病态工况是指因初始条件和操作条件改变而导致粗选产品质量不能满足后续浮选流程要求的工况.可以用粗选首槽泡沫品位作为粗选过程工况的评价指标.粗选过程中病态工况的识别是基于机器视觉的浮选过程监控的关键.通过长期观察发现,铜粗选槽溢流口处的泡沫状态能很好地反应泡沫上矿物附着的情况.如果目标矿物附着不好,泡沫顶部或在泡沫连接处因为没有承载金属矿粒呈现水化的反光区域,颜色为黑色.这一区域过大则水化现象严重,泡沫上附着金属矿粒较少,泡沫品位低;反之则泡沫坍塌现象严重,泡沫上附着的金属矿粒掉入矿浆,泡沫品位也会降低.黑色水化区域的大小能很好地反映当前浮选工况.某铜浮选粗选首槽泡沫图像及局部黑色水化区域如图2所示,对比泡沫背景,水化区域在视觉上呈现为黑色,与泡沫图像中目标矿物的颜色不一致,形状、大小没有规则.
2基于MPCA的图像局部光谱特征提取
多元图像分析是指利用PCA,PLS等多元统计分析工具,将多个通道图像数据投影到互不相关的主成分空间上,利用主元和图像像素变量之间的关系来提取图像特征[12-13].
将一帧原始RGB图像表示为一组由单变量组成的三维数据集合(I×J×M),其中I,J为像素几何坐标,M为光谱坐标,如图3所示.(I×J×M)可看作单变量图像fM(x,y)在M方向的堆叠,M=R,G,B.
先将(I×J×M)展开成2维数据矩阵X(N×M),如图3所示,其中N=I×J.于是,I×J个像素的fM(x,y)可以按照行或者列特定的顺序展开成一维的N×1图像像素矢量.
展开后的2维多元图像矩阵可以写成:
X(I×J)×M=[X1 X2 … XM]N×M.(1)
对X(N×M)进行PCA,将其分解成A(A≤M)个主成分的线性组合:
XN×M=∑Aa=1tapTa+E. (2)
式中:ta(a=1,2,…,A,A≤M)为标准正交的N维主成分得分矢量;pa(a=1,2,…,A,A≤M)为标准正交的M维主成分负载矢量;E为N×M维的残差矩阵.当A=M时,残差矩阵E为0矩阵.
对于展开后的多元图像矩阵X(N×M),一般有N远大于M,也就是矩阵X(N×M)在行方向上元素很多,在列方向上元素很少.对于这样的“窄”矩阵进行PCA分解,常采用构造“核”矩阵的方法[14-15]来减少计算时间.构造核矩阵:
K=XTX. (3)
其中K为Μ×Μ的低维核矩阵.
然后对K进行奇异值特征分解,得到的特征矢量就是主成分负载矢量pa,将pa根据特征值大小按照降序排列,得到排序以后的负载矢量pda,pd1为最大的特征值对应的特征矢量.由负载矢量pda,可计算出主元得分矢量tda:
tda=Xpda. (4)
得分矢量tda中的每个元素对应于3个变量(R,G,B)的加权平均像素,是不同像素位置的像素强度信息的压缩表述,代表了原图像中不同像素位置的光谱信息[16-18].如果同一图像中不同像素位置像素光谱特征相同,这些像素的得分值的关系将完全相同,即原始图像中所有具有相同光谱特征的像素的得分值在散点图中将重叠或者至少在同一区域.因此,根据累计贡献率选取主元个数,画出不同主元的得分矢量强度散点图并在散点图中标记出感兴趣的区域就可以捕获原始图像中的局部区域光谱特性.
依据公式(5),式中为Kronecker积,构建第一得分图像Ta (既d=1时的图像):
Ta=Xpda. (5)
然后利用得分值和组成该区域的像素变量之间的关系将标记的感兴趣区域映射到第一得分图像Ta上.
将特征像素值约束为0到255之间的整数,即:
Ta(i,j)=
RoundTa(i,j)-min [Ta(i,j)]max [Ta(i,j)]-min [Ta(i,j)]×255.(6)
式中:max [Ta(i,j)],min [Ta(i,j)]分别为主元图像中最大像素值和最小像素值.
统计像素点个数,计算标记区域的面积大小SL作为图像的局部区域光谱特征:
SL=N×Si. (7)
式中:N为标记区域的像素点个数;Si为单位像素面积.
3实验结果与分析
3.1基于MPCA的铜浮选泡沫图像局部光谱特征
提取算法
根据铜浮选泡沫图像中的黑色水化区域的特点,提出基于MPCA的铜浮选泡沫局部特征提取算法,其步骤如下:
1)将原始图像(I×J×M)展开成二维数据X(N×M),其中N=I×J.
2)构造核矩阵K=XTX,并对K矩阵进行奇异值分解,计算负载矢量pa,并将pa根据特征值大小按照降序排列,得到排序以后的负载矢量pda.
3)按式(4)计算主元得分矢量,计算累积贡献率CCR,根据累计贡献率CCR≥85%,选取主元个数.
4)依据选取的主元,绘制主元得分矢量强度散点图,标记局部区域对应的得分值聚集区或离群区,同时记录得分值所对应的特征像素值和空间位置.
5)按照式(5)重构第一得分图像,利用得分值和局部区域特征像素变量之间的关系,将得分矢量强度散点图中标记的区域映射到第一得分图像.
6)按照式(6)将特征像素值约束为0~255之间的整数.按照式(7)计算标记区域面积作为光谱特征.
3.2铜浮选泡沫图像采集及局部光谱特征提取
如图4所示,在铜粗选首槽泡沫表层上方110 cm处搭建浮选泡沫图像采集系统,系统由光源,工业摄像机,信号传输装置构成.摄像机视场范围为23.84 cm×17.88 cm,在如表1所示入矿条件下,采集泡沫图像样本200个,选取包含了明显黑色水化区域的典型图像作为训练图像,其余图像作为测试图像.同时采集对应时刻的铜粗选首槽泡沫样本,获得泡沫品位化验值.
针对训练图像,按照3.1节算法步骤1),2)建立MPCA全局模型,即计算负载矢量:
pd1=0.575 5 0.579 0.577 5T,
pd2=-0.523 0.117 0.845T.
然后按照步骤3),4)计算出测试图像的得分矢量,选取两个主元t1,t2,画出其得分矢量强度散点图,如图5所示,第一得分矢量值为-4~-2,第二得分矢量值为0.2~-0.6所对应的像素为局部区域特征像素.依据得分值和特征像素之间的关系,记录特征像素值、特征像素个数和空间位置.
根据步骤5)重构第一得分图像,利用得分值和特征像素变量之间的关系,将得分矢量强度散点图中对应的局部区域投影回第一得分图像,如图6所示.这一投影过程需结合图5,反复调整,直至所标记区域满意为止.统计特征像素个数,并根据单位像素面积,计算标记区域的面积SL.
最后,针对其余图像,重复步骤3)至步骤6),计算黑色水化区域的大小作为局部光谱特征:
SL=SL1,SL2,…,SL199,SL200.
3.3基于局部光谱特征的铜粗选“病态”工况识别
铜粗选过程是整个铜浮选流程的第1步,浮选性能好坏直接影响后续流程的操作和产品质量,通常用粗选首槽泡沫品位作为衡量粗选过程浮选性能的指标.将本文所提方法应用于泡沫图像样本,提取局部光谱特征,画出局部光谱特征与首槽泡沫品位的散点图,如图7所示.局部区域面积为15~28 cm2时对应的泡沫品位较高.当局部区域面积过大时(局部局域面积大于28 cm2),泡沫水化现象严重,泡沫上附着的金属矿粒少,泡沫品位低;而局部区域面积过小时(局部区域面积小于15 cm2),泡沫坍塌现象严重,泡沫上附着的金属矿粒掉入矿浆,泡沫品位也会降低.因此,由图7可知,便可以得到铜粗选首槽“病态”工况所对应的局部光谱特征阈值区间,识别出粗选过程的“病态”工况.
4工业应用
为了验证本文所提的方法,基于Visual C++和Matlab7.0开发了如图4所示的铜浮选泡沫图像监控系统应用于国内某铜浮选厂粗选流程.该系统能够提供浮选泡沫视觉图像和对应的图像视觉特征曲线,实现了铜浮选粗选首槽病态工况的识别,并将其总结为专家控制规则,现场工作人员能及时了解工况,根据工况的变化调整操作以稳定和提高浮选品位及回收率.2012年1-5月,所开发系统在工业现场连续试运行5个月,分析对比入矿条件基本相同,药剂制度相同情况下的2010年回收率数据,如图8所示,系统投入运行前铜回收率平均值为86.48%,标准差0.846 759;投入运行后铜回收率平均值为87.23% ,标准差为0.825 57.在一定程度上,对于稳定和提高铜回收率指标有帮助.
5结论
铜粗选工况识别是铜浮选全流程监控的关键.本文描述了铜浮选粗选过程,分析了影响粗选过程的主要因素和粗选首槽泡沫图像黑色水化区域形成机理,提出以黑色水化区域面积作为铜浮选粗选过程病态工况识别的局部图像特征,并针对这一特征大小、形状无规则性,提出一种基于MPCA的局部光谱特征提取新方法.该方法无需考虑原始图像中的像素空间位置,能很好地捕获原始图像的局部光谱特征.所提取的特征与浮选泡沫品位有很强的相关性,可用于铜浮选粗选过程病态工况识别.但是入矿类型的改变会引起粗选工况区间的漂移,用单一的图像特征会造成病态工况区间的误识别.这将是我们下一步要解决的问题.
参考文献
[1]桂卫华.矿物浮选泡沫图像处理与过程监测技术[M]. 长沙:中南大学出版社,2013:11-17.
GUI Weihua. Mineral flotation froth image processing and process monitoring technology[M].Changsha: Central South University Press, 2013.(In Chinese)
[2]桂卫华,阳春华,卢明,等.基于机器视觉的矿物浮选过程监控技术研究进展[J].自动化学报,2013, 39(11):1879-1887.
GUI Weihua, YANG Chunhua, LU Ming, et al. Machinevisionbased online measuring and controlling technologies for mineral flotation—a review[J]. Acta Automatica Sinica, 2013, 39(11):1879-1887. (In Chinese)
[3]CIT IR C, AKTAS Z, BER BERR. Off line image analysis for froth flotation of coal[J]. Computers & Chemical Engineering, 2004,28(60):625-632.
[4]HATONEN J. Image analysis in mineral flotation[D]. Helsinki, Finland: Helsinki University of Technology, 1999.
[5]ALDRICH C, MAIIAIS C, SHEAN B J, et al. Online monitoring and control of froth flotation systems with machine vision:a review[J]. International Journal of Mineral Processing, 2010,96(4):1-13.
[6]XU C H, GUI W H, YANG C H. Flotation process fault detection using output PDF of bubble size distribution[J]. Minerals Engineering, 2012,26:5-12.
[7]BONIFAZI G, SERRANTI S,VOLPE F,et al.Characterization of flotation froth colour and structure by machine vision[J]. Computers & Geosciences, 2001,27(9):1111-1117.
[8]BONIFAZI G,SERRANTI S,VOLPE F,et al.Flotation froth characterization by closed domain (bubbles) color analysis[C]//4th Int Conf on Quality Control by Artificial Vision,November 10-12, Takamatsu, Japan, 1998:131-137.
[9]BARTOLACCI G,PELLETIER R,TESSIER J.Application of numerical image analysis to process diagnosis and physical parameter measurement in mineral processes —part 1: flotation control based on froth textural characteristic[J]. Minerals Engineering, 2006,19(6/8):734-747.
[10]KAARTINEN J. Data acquisition and analysis system for mineral flotation[D].Finland:Control Engineering Laboratory, Helsinki University of Technology, 2001.
[11]KAARTINEN J, HATONEN J, MIETTUNEN J, et al. Image analysis based control of zinc flotation a multicamera approach[C]//Preprints of the Seventh International Conference on Control, Automation, Robotics and Vision(ICARV 2002), Singapore,2002.
[12]GARCAMUOZ S,GIERER D S.Coating assessment for colored immediate release tablets using multivariate image analysis[J].International Journal of Pharmaceutics,2010,395:104-113.
[13]ESBENSEN K,GELADI P.Strategy of multivariate image analysis[J].Chemometrics and Intelligent Laboratory Systems,1989,7(1/2):67-86.
[14]PARTSMONTALBAN J M,DE JUAN A,FERRER A. Multivariate image analysis: a review with applications[J]. Chemometrics and Intelligent Laboratory Systems,2011,107(1):1-23.
[15]GELADI P,ISAKSSON H,LINDQVIST L,et al.Principal component analysis of multivariate images[J].Chemometrics and Intelligent Laboratory Systems,1989,5(3):209-220.
[16]YU H,MACGREGOR J F.Multivariate image analysis and regression for prediction of coating and distribution in the production of snack foods[J].Chemometrics and Intelligent Laboratory Systems,2003,67(2):125-144.
[17]LIU J J,MACGREGOR J F.Estimation and monitoring of product aesthetics: application to manufacturing of “engineered stone” countertops[J].Machine Vision and Applications, 2006,16:374-383.
[18]许悟生,谢柯夫.基于像素灰度关联的边缘检测[J].湖南师范大学自然科学学报,2012,35(4):26-30.
XU Wusheng, XIE Kefu. Edge detection based on pixel gray correlation[J].Journal of Natural Science of Hunan Normal University,2012,35(4):26-30.(In Chinese)
[3]CIT IR C, AKTAS Z, BER BERR. Off line image analysis for froth flotation of coal[J]. Computers & Chemical Engineering, 2004,28(60):625-632.
[4]HATONEN J. Image analysis in mineral flotation[D]. Helsinki, Finland: Helsinki University of Technology, 1999.
[5]ALDRICH C, MAIIAIS C, SHEAN B J, et al. Online monitoring and control of froth flotation systems with machine vision:a review[J]. International Journal of Mineral Processing, 2010,96(4):1-13.
[6]XU C H, GUI W H, YANG C H. Flotation process fault detection using output PDF of bubble size distribution[J]. Minerals Engineering, 2012,26:5-12.
[7]BONIFAZI G, SERRANTI S,VOLPE F,et al.Characterization of flotation froth colour and structure by machine vision[J]. Computers & Geosciences, 2001,27(9):1111-1117.
[8]BONIFAZI G,SERRANTI S,VOLPE F,et al.Flotation froth characterization by closed domain (bubbles) color analysis[C]//4th Int Conf on Quality Control by Artificial Vision,November 10-12, Takamatsu, Japan, 1998:131-137.
[9]BARTOLACCI G,PELLETIER R,TESSIER J.Application of numerical image analysis to process diagnosis and physical parameter measurement in mineral processes —part 1: flotation control based on froth textural characteristic[J]. Minerals Engineering, 2006,19(6/8):734-747.
[10]KAARTINEN J. Data acquisition and analysis system for mineral flotation[D].Finland:Control Engineering Laboratory, Helsinki University of Technology, 2001.
[11]KAARTINEN J, HATONEN J, MIETTUNEN J, et al. Image analysis based control of zinc flotation a multicamera approach[C]//Preprints of the Seventh International Conference on Control, Automation, Robotics and Vision(ICARV 2002), Singapore,2002.
[12]GARCAMUOZ S,GIERER D S.Coating assessment for colored immediate release tablets using multivariate image analysis[J].International Journal of Pharmaceutics,2010,395:104-113.
[13]ESBENSEN K,GELADI P.Strategy of multivariate image analysis[J].Chemometrics and Intelligent Laboratory Systems,1989,7(1/2):67-86.
[14]PARTSMONTALBAN J M,DE JUAN A,FERRER A. Multivariate image analysis: a review with applications[J]. Chemometrics and Intelligent Laboratory Systems,2011,107(1):1-23.
[15]GELADI P,ISAKSSON H,LINDQVIST L,et al.Principal component analysis of multivariate images[J].Chemometrics and Intelligent Laboratory Systems,1989,5(3):209-220.
[16]YU H,MACGREGOR J F.Multivariate image analysis and regression for prediction of coating and distribution in the production of snack foods[J].Chemometrics and Intelligent Laboratory Systems,2003,67(2):125-144.
[17]LIU J J,MACGREGOR J F.Estimation and monitoring of product aesthetics: application to manufacturing of “engineered stone” countertops[J].Machine Vision and Applications, 2006,16:374-383.
[18]许悟生,谢柯夫.基于像素灰度关联的边缘检测[J].湖南师范大学自然科学学报,2012,35(4):26-30.
XU Wusheng, XIE Kefu. Edge detection based on pixel gray correlation[J].Journal of Natural Science of Hunan Normal University,2012,35(4):26-30.(In Chinese)
[3]CIT IR C, AKTAS Z, BER BERR. Off line image analysis for froth flotation of coal[J]. Computers & Chemical Engineering, 2004,28(60):625-632.
[4]HATONEN J. Image analysis in mineral flotation[D]. Helsinki, Finland: Helsinki University of Technology, 1999.
[5]ALDRICH C, MAIIAIS C, SHEAN B J, et al. Online monitoring and control of froth flotation systems with machine vision:a review[J]. International Journal of Mineral Processing, 2010,96(4):1-13.
[6]XU C H, GUI W H, YANG C H. Flotation process fault detection using output PDF of bubble size distribution[J]. Minerals Engineering, 2012,26:5-12.
[7]BONIFAZI G, SERRANTI S,VOLPE F,et al.Characterization of flotation froth colour and structure by machine vision[J]. Computers & Geosciences, 2001,27(9):1111-1117.
[8]BONIFAZI G,SERRANTI S,VOLPE F,et al.Flotation froth characterization by closed domain (bubbles) color analysis[C]//4th Int Conf on Quality Control by Artificial Vision,November 10-12, Takamatsu, Japan, 1998:131-137.
[9]BARTOLACCI G,PELLETIER R,TESSIER J.Application of numerical image analysis to process diagnosis and physical parameter measurement in mineral processes —part 1: flotation control based on froth textural characteristic[J]. Minerals Engineering, 2006,19(6/8):734-747.
[10]KAARTINEN J. Data acquisition and analysis system for mineral flotation[D].Finland:Control Engineering Laboratory, Helsinki University of Technology, 2001.
[11]KAARTINEN J, HATONEN J, MIETTUNEN J, et al. Image analysis based control of zinc flotation a multicamera approach[C]//Preprints of the Seventh International Conference on Control, Automation, Robotics and Vision(ICARV 2002), Singapore,2002.
[12]GARCAMUOZ S,GIERER D S.Coating assessment for colored immediate release tablets using multivariate image analysis[J].International Journal of Pharmaceutics,2010,395:104-113.
[13]ESBENSEN K,GELADI P.Strategy of multivariate image analysis[J].Chemometrics and Intelligent Laboratory Systems,1989,7(1/2):67-86.
[14]PARTSMONTALBAN J M,DE JUAN A,FERRER A. Multivariate image analysis: a review with applications[J]. Chemometrics and Intelligent Laboratory Systems,2011,107(1):1-23.
[15]GELADI P,ISAKSSON H,LINDQVIST L,et al.Principal component analysis of multivariate images[J].Chemometrics and Intelligent Laboratory Systems,1989,5(3):209-220.
[16]YU H,MACGREGOR J F.Multivariate image analysis and regression for prediction of coating and distribution in the production of snack foods[J].Chemometrics and Intelligent Laboratory Systems,2003,67(2):125-144.
[17]LIU J J,MACGREGOR J F.Estimation and monitoring of product aesthetics: application to manufacturing of “engineered stone” countertops[J].Machine Vision and Applications, 2006,16:374-383.
[18]许悟生,谢柯夫.基于像素灰度关联的边缘检测[J].湖南师范大学自然科学学报,2012,35(4):26-30.
XU Wusheng, XIE Kefu. Edge detection based on pixel gray correlation[J].Journal of Natural Science of Hunan Normal University,2012,35(4):26-30.(In Chinese)