一种基于双能X射线的多通道图像融合铜矿分选方法  

A Multi-channel Image Fusion Method for Copper Ore Separation Based on Dual-energy X-ray

作　　者：余茳山 何剑锋[1,2] 李卫东[1,2] 聂逢君[1] 夏菲[1] 汪雪元[1,2] 袁兆林 瞿金辉 钟国韵[1,2] YU Jiangshan;HE Jianfeng;LI Weidong;NIE Fengjun;XIA Fei;WANG Xueyuan;YUAN Zhaolin;QU Jinhui;ZHONG Guoyun(Jiangxi Engineering Technology Research Center of Nuclear Geoscience Data Science and System,East China University of Technology,Nanchang 330013,China;Information Engineering College,East China University of Technology,Nanchang 330013,China)

机构地区：[1]东华理工大学江西省核地学数据科学与系统工程技术研究中心,南昌330013 [2]东华理工大学信息工程学院,南昌330013

出　　处：《有色金属(中英文)》2025年第2期275-280,共6页Nonferrous Metals

基　　金：国家自然科学基金资助项目(U2067202);江西省主要学科学术和技术带头人培养计划(20225BCJ22004)。

摘　　要：传统方法分析单张R值图像或T值图像时信息不够全面,高、低能图像难以提取特征且易受噪声干扰,限制了识别效果;R值图像特征提取相对容易且抗干扰能力强,但缺少厚度信息;T值图像保留了厚度和密度信息,但在复杂成分识别方面效果不佳且对比度较低。为了改善上述问题,提出了一种将高、低能图像融合后与R值图像、T值图像进行通道合并的多通道图像融合的方法。合并后的图像综合了高、低能、R值和T值图像信息,弥补各自信息的不足,达到更全面的效果。实验对两类品位不同的铜矿分选,用ResNet18对几种不同处理方式的图像训练测试,测试结果为双能融合图像与R值图像和T值图像通道合并形成的双能融合RT合并图像的分选准确率最高,达到95.80%。Traditional methods for analyzing single R-value or T-value images lack comprehensive information,making it difficult to extract features from high and low energy images and susceptible to noise interference,which limits recognition effectiveness.The feature extraction of R-value images is relatively easy and has strong anti-interference ability,but lacks thickness information.The T-value image preserves thickness and density information,but performs poorly in complex component recognition and has low contrast.To improve the above issues,a multi-channel image fusion method that combines high and low energy images with R-value images and T-value images for channel merging was proposed.The merged image combines the image information of high energy,low energy,R-value and T-value to compensate for the shortcomings of their respective information and achieve a more comprehensive effect.The experiment focused on the separation of two types of copper mines with different grades.ResNet18 was used to train and test the images of several different processing methods.The test result showed that the separation accuracy of the dual energy fusion RT merged image formed by combining the dual energy fusion image with R-value image and T-value image channels was the highest,reaching 95.80%.

关 键 词：铜矿分选 通道合并 图像融合 R值图像 T值图像 

分 类 号：TD952[矿业工程—选矿]