基于图像特征提取的煤浆灰分预测模型改进设计

Improved design of coal slurry ash prediction model based on image feature extraction

针对煤浆灰分测量难度大、预测精度低的问题,提出了一种基于图像特征提取的煤浆灰分预测模型的改进设计方案。该方法分别从图像灰度均值、方差、能量等六个指标中提取煤浆的特征向量;分析BP神经网络预测模型的流程,引入高斯函数改进该模型,构建径向基神经网络,将特征向量作为网络输入参量,经过初始化、输出值运算和误差修正等过程,获得最终预测输出矩阵,完成预测模型的改进设计。仿真实验结果表明,改进模型的灰分预测结果与实际结果更接近,其误差更小。

Aiming at the problems of difficult measurement and low prediction accuracy of coal slurry ash content,an improved design scheme of coal slurry ash prediction model based on the image feature extraction was proposed.The method respectively extracted the feature vectors of the coal slurry from six indexes,such as image gray value,variance,energy,etc.The process of BP neural network prediction model was analyzed,Gaussian function was introduced to improve the model,the radial basis function neural network was constructed,the feature vectors were taken as the network input parameter,the final prediction output matrix was obtained through the process of initialization,output value operation and error correction,and the improved design of prediction model was completed.The simulation results showed that the ash prediction results of the improved model were closer to the actual results,with less error.