基于机器视觉的浮选泡沫灰分在线检测研究

Research on On-line Detection of Flotation Froth Ash Based on Machine Vision

通过实验室搭建的煤泥浮选泡沫采集系统对不同浮选阶段的泡沫图像进行采集和预处理,分析了不同阶段的泡沫图像的特征。分别基于图像灰度直方图和灰度共生矩阵方法提取泡沫纹理特征参数,针对泡沫分割过程中的过分割和欠分割问题,建立基于自适应标记提取的改进分水岭算法实现泡沫图像的有效分割,基于此提取了泡沫面积和承载量两个特征参数。综合泡沫纹理特征参数、尺寸特征参数及泡沫稳定性构建了图像多维特征向量,使特征敏感度得到一定程度的增强,在此基础上,构建了预测精煤灰分的BP神经网络拓扑结构。精煤灰分预测结果表明:当BP神经网络模型的隐含层节点数为10时,灰分在线检测的绝对误差可以控制在±1%之内。

This paper collects and preprocesses the foam images of different flotation stages through the slime flotation foam collection system built in the laboratory,and analyzes the characteristics of the foam images at different stages,based on the gray-level histogram and gray-level co-occurrence matrix methods.The foam texture feature parameters are extracted.Aiming at the over-segmentation and under-segmentation problems in the foam segmentation process,an improved watershed algorithm based on adaptive marker extraction is established to achieve effective segmentation of foam images.Based on this,two feature parameters of foam area and carrying capacity are extracted.Combining foam texture feature parameters,size feature parameters and foam stability,a multi-dimensional image feature vector is constructed to enhance feature sensitivity to a certain extent.On this basis,a BP neural network topology structure for predicting the ash content of clean coal is constructed.The number of hidden layer nodes of BP neural network model reaches 10,the absolute error of the ash online detection can be controlled within±1%,It has valuable reference and significance for the actual coal preparation plant to realize the automatic control in the flotation process.