摘要
针对矿物浮选过程中回收率难以在线检测的问题,提出了一种回收率预测方法。采用最小二乘支持向量机构造预测模型,以图像特征作为模型输入,通过交叉验证实现模型参数优化。为提取泡沫特征,通过计算图像相对红色分量提取颜色特征,结合聚类与分水岭方法分割泡沫图像并提取尺寸特征,利用像素分析方法提取承载量特征,采用图像对的相关性分析方法提取泡沫速度、破碎率等动态特征,并对泡沫特征与回收率进行了相关性分析。实验结果表明,该方法能有效预测回收率。
Aiming at the difficulty of online measurement of mineral recovery in mineral flotation process, a recovery prediction method based on froth features extraction is proposed in this paper. The prediction model is built by least squares support vector machine （LS-SVM） with image features as input, and the regulating parameters are optimized by a cross validation method. In order to extract froth image features, the froth color feature is extracted by using computing relative redness component, and the bubble size is measured by combining clustering and watershed segmentation method. Using pixel analysis method the bubble load is calculated, and the dynamic features such as froth velocity and bubble collapse are extracted by analyzing correlation of image pairs. To verify the relationship between froth features and mineral recovery, the correlation analyses are also carried out. The experimental results demonstrate the proposed method can predict floatation mineral recovery effectively.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2009年第6期1295-1300,共6页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金重点项目(60634020)
国家自然科学基金(60874069
60804037)
国家863项目(2006AA04Z181)
关键词
矿物浮选
泡沫图像
特征提取
回收率预测
最小二乘支持向量机
mineral flotation
froth image
feature extraction
recovery prediction
least squares support vector machine （ LS-SVM ）
