基于支持向量机的纹理特征识别方法研究  

Research on texture feature recognition based on support vector machines

作　　者：梁丽媛 陈永当[1] 胡梦麒 李婵 苗杨杨 LIANG Liyuan;CHEN Yongdang;HU Mengqi;LI Chan;MIAO Yangyang(School of Mechanical and Electrical Engineering,Xi′an Polytechnic University,Xi′an 710600,China)

机构地区：[1]西安工程大学机电工程学院,西安710600

出　　处：《智能计算机与应用》2024年第4期168-172,共5页Intelligent Computer and Applications

基　　金：陕西省重点研发项目(2022GY-307)。

摘　　要：触觉能力是使人类能够与外界环境进行交互的一项非常重要的能力,为了模仿该能力并且使其能够应用在仿生学研究中,本文采用薄膜式压力传感器,搭建了并列式触觉传感阵列,使用多通路信号采集方法,研究薄膜式触觉传感器在接触不同麻将牌面时的纹理识别情况。在此基础上,本文基于Labview编写了薄膜触觉传感器的多通路数据采集程序,通过对所采集的数据进行拟合处理,验证了以麻将牌面为对象的特征分类识别方法。最后,采用支持向量机分类算法,以实验平台采集的数据为训练样本,建立了特征分类模型和特征识别模型,并对不同核函数对模型的性能影响进行了探究。The sense of touch is a very important ability that enables humans to interact with the external environment.In order to mimic this ability and apply it to bionics research,a thin-film pressure sensor was developed to build a parallel touch sensor array in this study.A multi-channel signal acquisition method was employed to investigate the texture recognition of thin-film touch sensors when in contact with different Mahjong tiles.With that said,a multi-channel data acquisition program for thin-film touch sensors was developed using Labview.Through fitting and processing the collected data,a feature classification recognition method for Mahjong tiles was validated.Finally,a support vector machine classification algorithm was used to establish feature classification and models recognition using the data collected from the experimental platform as training samples,and the influence of different kernel functions on the performance of the model is explored.

关 键 词：纹理识别 传感器 触觉感知 支持向量机 

分 类 号：TP241.2[自动化与计算机技术—检测技术与自动化装置]