改进LR算法在矿浆筛分水浊度分类中的应用

Application of Improved LR Algorithm in the Turbidity Classification of Mineral Pulp Screening Water

针对传统矿浆细度检测方法离线筛分法的矿浆筛分水浊度判别问题,为开发出细度自动检测技术,提出改进LR(Logistic Regression,逻辑回归)算法,对矿浆筛分水进行浊度分类。首先,采集矿浆筛分水图像,并采用矿浆浊度信息提取算法得到筛分水图像的浊度信息数据集;其次,分析传统LR算法的不足,提出一种改进的LR算法;最后,将改进的LR算法运用于浊度信息数据集的分类。结果表明,采用基于RGB的矿浆筛分水图像浊度信息提取算法提取矿浆筛分水图像特征参数,利用L2正则化改进逻辑回归算法后,其指标较其他算法均得到优化,对矿浆筛分水浊度分类问题具有良好的分类效果。

Aiming at the problem of turbidity discrimination of mineral pulp screening water in the offline screening method of traditional mineral pulp fineness detection, in order to develop an automatic fineness detection technology, an improved LR(Logistic Regression) algorithm was proposed to classify the turbidity of mineral pulp screening water. Firstly, the images of mineral pulp screening water were collected, and the turbidity information data sets of screening water images were obtained by using the turbidity information extraction algorithm. Secondly, an improved LR algorithm was proposed by analyzing the shortcomings of the traditional LR algorithm. Finally, the improved LR algorithm was applied to the classification of turbidity information data sets. The results show that after using turbidity information extraction algorithm based on RGB to extract the characteristic parameters of mineral pulp screening water images, as well as improving the logistic regression algorithm by L2 regularization, the indexes were optimized compared with other algorithms, achieving a better classification effect on the turbidity classification of mineral mineral pulp screening water.