一种基于图像特征提取的浮选回收率预测算法  

A flotation recovery prediction algorithm based on image feature extraction

作　　者：周开军[1] 阳春华[1] 牟学民[1] 桂卫华[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《高技术通讯》2009年第9期957-963,共7页Chinese High Technology Letters

基　　金：国家自然科学基金(60634020;60874069;60804037)资助项目。

摘　　要：针对矿物浮选过程中的一类回收率预测问题,提出了一种基于泡沫图像特征提取的预测算法。该算法采用最小二乘支持向量机(LSSVM)建立预测模型,通过施密特正交化对核矩阵进行简约,利用核偏最小二乘方法(KPLS)进行LSSVM参数辨识,以此构造具有稀疏性的LSSVM,有效地减小了算法的计算复杂度。为检验模型泛化及预测能力,为多个泡沫特征信息引入预测模型,采用泡沫图像特征提取方法提取泡沫颜色、速度、尺寸、承载量及破碎率特征。实验结果表明,该预测算法对浮选回收率具有良好预测效果。Aiming at a class of mineral recovery prediction problems in mineral flotation, the paper presents a prediction algorithm based on froth image feature extraction. The prediction model is built by using the least squares support vector machine （LSSVM）, and the kernel matrix is reduced by the Schmidt orthogonalization to obtain the base vectors of kernel matrix. In order to obtain the LSSVM with a sparse property, the LSSVM parameters are identified by the kernel partial least squares （KPLS）, and the algorithm computation complexity is decreased effectively. For the purpose of verifying the generalization and prediction performance, multiple froth characteristic information are incorporated into the prediction model. Furthermore, the image features such as froth color, velocity, bubble size, bubble loading and collapse rate are extracted through froth image feature extraction methods. The experimental results show that the proposed algorithm performs well on flotation recovery prediction.

关 键 词：矿物浮选 泡沫图像 预测模型 最小二乘支持向量机(LSSVM) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]