基于分段T值曲线拟合的双能X射线的铜矿分选研究

Research on Copper Ore Sorting Using Dual-Energy X-Ray Technique Based on Curve Fitting of T-value by Segmented Straight Lines

针对传统T值曲线拟合算法的双能X射线透射技术分选铜矿偏差较大的问题,提出了一种分段T值曲线拟合方法,该方法将厚度进行区间划分,在每个区间内进行一次T值曲线拟合来对铜矿进行分选,能更准确体现不同厚度下T值的变化趋势,减小厚度给分选带来的影响。采用高能、低能射线拟合T值曲线,再在各个区间内计算T值与曲线拟合映射值的残差,不同物质残差也存在不同,以此达到分选铜矿的目的。验证实验结果表明,对两类不同品位的铜矿进行分选,将经过分段T值曲线拟合的矿石图像放入ResNet18神经网络中进行训练,并用训练好的模型进行测试,测试准确率达到88.67%。

Copper ore sorting by dual-energy X-ray technique basded on traditional algorithms by curve fitting of T-value always results in great deviation.In view of this problem,a curve fitting of T-value by segmented straight lines was proposed.This method divides thickness into several segments,and performs a T-value curve fitting in each segment for copper ore sorting,which can more accurately reflect the changing trend of T-value with different thicknesses and also reduce impact of thickness on sorting.High-energy and low-energy rays are adopted to fit the T-value curve,and then the residual between the T-value and the curve-fitting mapping value is calculated in each segment,showing different residuals for different substances,which can be used to achieve copper ore sorting.In a verification experiment,two types of copper ores with different grades were sorted by this method.The ore images fitted with T-value curves by segmented straight lines were put into ResNet18 neural network for training,and the trained model was then used for testing.The test results show that an accuracy rate can reach 88.67%.