基于泡沫纹理信度分配SVM的矿物浮选工况识别.pdfVIP

学兔兔 第32卷 第10期 仪 器 仪 表 学 报 VoI．32 No．10 2011年 10月 Chinese Journal of Scientific Instrument 0ct．2011 基于泡沫纹理信度分配SVM的矿物浮选工况识别术 阳春华，任会峰，桂卫华，鄢 锋，唐朝晖 (中南大学信息科学与工程学院 长沙 410083) 摘 要：针对矿物浮选过程中泡沫纹理各向异性、纹理特征信度差异以及类间样本数分布不均等问题，提出一种浮选工况的自 动识别方法。首先基于多角度融合的空间灰度共生矩阵计算角二阶矩、熵、惯性矩、逆差矩和相关性等二阶统计量描述泡沫纹 理特征，然后引入信息熵的概念，以二阶统计量的信息增益分配各种纹理特征对分类器的信度，再通过类间样本数加权策略消 除样本分布的不平衡性，最后实现了一对一加权支持向量机的浮选工况识别。工业现场泡沫图像测试结果表明，该方法能够有 效降低小样本的错分率，提高整体识别率。 关键词：泡沫浮选；工况识别；纹理特征；信度分配；支持向量机 中图分类号：TP274．3 文献标识码：A 国家标准学科分类代码：51O．4O Performance recognition using texture credit distributed SVM for froth flotation process Yang Chunhua，Ren Huifeng，Gui Weihua，Yan Feng，Tang Zhaohui (School of Information Science and Engineering，Central South University，Changsha 410083，China) Abstract：A performance recognition method for mineral froth flotation process is presented，where multiple imbal— ance exists in texture directions，feature credit and sample distribution．Firstly，second order statistics such as angu— lar second moment，entropy，moment of inertia，moment of deficit and relevance are extracted based on four-direc— tion fused spatial gray co—occurrence matrix to avoid the adverse effects of directions．Then information entropy is in— troduced to evaluate the feature credit of the texture features via the information gains．Furthermore，the imbalance of sample distribution is eliminated with between—class weighted strategy，and a performance recognition model is estab— lished using one—versus-one weighted support vector machine(OVO—WSVM)．Industrial on—site experiment verifies that the proposed method can decrease the misclassification rate of small samples and increase overall recognition rate．