激光发生器中的数据分类数学模型设计
|
摘要
针对激光发生器中分类方法的收敛性较差的问题。本文依据随机森林法,对激光发生器中的数据分类数学模型进行设计,利用统计去噪法对激光发生器中的数据进行去噪,对数据点密度参数特征,在设定阈值范围内部还是外部进行判断,并将范围外部的数据点确定为噪声数据点去除,为高精度的数据分类奠定基础;依据3D-HT特征提取法对激光发生器中的数据特征进行提取,通过激光发生器中的数据原本结构,迅速恢复全部数据的结构,和数据邻近关系,以估计各数据点法矢特征向量;利用前向法对所得特征向量进行筛选,获得特征空间,以增强分类方法的收敛性;根据自顶而下的贪婪法,使得决策树各内部结点能够选取数据分类最佳的特征属性,并将达到本结点的数据区分为两类或者是多类,对这个过程进行重复,一直到可以精确地将全部的训练数据聚类完毕。实验表明,所提方法提高了激光发生器中数据分类的精度和收敛性,相比当前方法具有的一定的优越性。
Aiming at the poor convergence for the classification method in the laser generator,the mathematical model of the data classification in the laser generator is designed based on random forest method. The data in the laser generator is denoised by statistical denoising method. Judge the data point density parameter characteristics whether it's inside or outside within the set threshold range and the data points outside the range are determined as noise data point removal,which lays a foundation for high-precision data classification. The data features in the laser generator are extracted according to the 3D-HT feature extraction method. The original structure of the data quickly restores the structure of all data and the data neighbor relationship to estimate the normal vector of each data point. The forward feature method is used to filter the obtained feature vectors to obtain the feature space to enhance the convergence of the classification method. The top-down greedy method enables the internal nodes of the decision tree to select the best feature attributes of the data classification,and divides the data reaching the node into two or more categories,repeating this process until all training data can be accurately clustered. Experiments show that the proposed method improves the accuracy and convergence of data classification in the laser generator,and has certain advantages over the current method.
Aiming at the poor convergence for the classification method in the laser generator,the mathematical model of the data classification in the laser generator is designed based on random forest method. The data in the laser generator is denoised by statistical denoising method. Judge the data point density parameter characteristics whether it's inside or outside within the set threshold range and the data points outside the range are determined as noise data point removal,which lays a foundation for high-precision data classification. The data features in the laser generator are extracted according to the 3D-HT feature extraction method. The original structure of the data quickly restores the structure of all data and the data neighbor relationship to estimate the normal vector of each data point. The forward feature method is used to filter the obtained feature vectors to obtain the feature space to enhance the convergence of the classification method. The top-down greedy method enables the internal nodes of the decision tree to select the best feature attributes of the data classification,and divides the data reaching the node into two or more categories,repeating this process until all training data can be accurately clustered. Experiments show that the proposed method improves the accuracy and convergence of data classification in the laser generator,and has certain advantages over the current method.
引文
[1]陈晓,贾华宇,郭燕.基于单片机的激光发生器的驱动电路设计[J].应用光学,2015,36(1):134-139.
[2]刘志青,李鹏程,陈小卫,等.基于信息向量机的机载激光雷达点云数据分类[J].光学精密工程,2016,24(1):210-219.
[3]卢东,钟祝强,夏光琼,等.用双滤波反馈半导体激光器产生低延时特征的混沌信号[J].光子学报,2016,45(10):13-18.
[4]刘莉,杨傲雷,屠晓伟,等.面向INS数据分类的鲁棒性无监督聚类方法[J].仪器仪表学报,2016,37(1):152-160.
[5]刘永明,邓孺孺,秦雁,等.机载激光雷达测深数据处理与应用[J].遥感学报,2017,21(6):982-995.
[6]喻篧慎,云挺,杨绪兵.基于激光点云数据的植物器官多维特征分割方法[J].数据采集与处理,2015,30(5):1054-1061.
[7]李璞,江镭,孙媛媛,等.面向全光物理随机数发生器的混沌实时光采样研究[J].物理学报,2015,64(23):25-34.
[8]刘志青,李鹏程,陈小卫,等.基于信息向量机的机载激光雷达点云数据分类[J].光学精密工程,2016,24(1):210-219.
[9]孙倩倩,杜敏,郭连波,等.塑料激光诱导击穿光谱技术快速分类应用研究[J].光谱学与光谱分析,2017,37(7):2205-2209.
[10]梁迎磊,杨苏辉,赵长明,等.半导体抽运Nd:YAG激光器强度噪声抑制的研究[J].激光技术,2016,40(1):113-117.
[11]黄超,赵学庆,易爱平,等.电子束泵浦Xe Cl准分子激光器及应用[J].强激光与粒子束,2015,27(4):62-66.
[12]李荣雨,程磊.基于SVM最优决策面的决策树构造[J].电子测量与仪器学报,2016,30(3):342-351.
[13]梁栋,张凤琴,陈大武,等.一种基于决策树和遗传算法-BP神经网络的组合预测模型[J].中国科技论文,2015,10(2):169-174.
[14]谭造乐,郝志峰,蔡瑞初,等.基于信息增益的Hadoop瓶颈检测算法[J].电信科学,2016,32(7):115-120.
[15]曹杰,邵笑笑.基于信息增益和Bagging集成学习算法的个人信用评估模型研究[J].数学的实践与认识,2016,46(8):90-98.
[16]韩丽娜,韩改宁.决策树算法在学生成绩分析中的应用研究[J].电子设计工程,2017,25(2):18-21.
[2]刘志青,李鹏程,陈小卫,等.基于信息向量机的机载激光雷达点云数据分类[J].光学精密工程,2016,24(1):210-219.
[3]卢东,钟祝强,夏光琼,等.用双滤波反馈半导体激光器产生低延时特征的混沌信号[J].光子学报,2016,45(10):13-18.
[4]刘莉,杨傲雷,屠晓伟,等.面向INS数据分类的鲁棒性无监督聚类方法[J].仪器仪表学报,2016,37(1):152-160.
[5]刘永明,邓孺孺,秦雁,等.机载激光雷达测深数据处理与应用[J].遥感学报,2017,21(6):982-995.
[6]喻篧慎,云挺,杨绪兵.基于激光点云数据的植物器官多维特征分割方法[J].数据采集与处理,2015,30(5):1054-1061.
[7]李璞,江镭,孙媛媛,等.面向全光物理随机数发生器的混沌实时光采样研究[J].物理学报,2015,64(23):25-34.
[8]刘志青,李鹏程,陈小卫,等.基于信息向量机的机载激光雷达点云数据分类[J].光学精密工程,2016,24(1):210-219.
[9]孙倩倩,杜敏,郭连波,等.塑料激光诱导击穿光谱技术快速分类应用研究[J].光谱学与光谱分析,2017,37(7):2205-2209.
[10]梁迎磊,杨苏辉,赵长明,等.半导体抽运Nd:YAG激光器强度噪声抑制的研究[J].激光技术,2016,40(1):113-117.
[11]黄超,赵学庆,易爱平,等.电子束泵浦Xe Cl准分子激光器及应用[J].强激光与粒子束,2015,27(4):62-66.
[12]李荣雨,程磊.基于SVM最优决策面的决策树构造[J].电子测量与仪器学报,2016,30(3):342-351.
[13]梁栋,张凤琴,陈大武,等.一种基于决策树和遗传算法-BP神经网络的组合预测模型[J].中国科技论文,2015,10(2):169-174.
[14]谭造乐,郝志峰,蔡瑞初,等.基于信息增益的Hadoop瓶颈检测算法[J].电信科学,2016,32(7):115-120.
[15]曹杰,邵笑笑.基于信息增益和Bagging集成学习算法的个人信用评估模型研究[J].数学的实践与认识,2016,46(8):90-98.
[16]韩丽娜,韩改宁.决策树算法在学生成绩分析中的应用研究[J].电子设计工程,2017,25(2):18-21.