基于模糊支持向量机和D-S证据理论的钨矿石初选方法  

Ungsten Ore Primary Selection Based on Fuzzy Support Vector Machine and D-S Evidence Theory

作　　者：胡发焕[1,2] 刘国平[1] 胡瑢华[1] 董增文[1] 

机构地区：[1]南昌大学机电工程学院,南昌330031 [2]江西理工大学机电工程学院,江西赣州341000

出　　处：《光子学报》2017年第7期165-172,共8页Acta Photonica Sinica

基　　金：国家自然科学基金(No.71361014)资助~~

摘　　要：单一特征识别的钨矿石初选准确率低,稳定性差,本文提出结合模糊支持向量机和D-S证据理论相的多特征钨矿石识别方法.对矿石图像预处理后,分别提取矿石的颜色、灰度和纹理等3类视觉特征,对这3类视觉特征进行模糊分类得到各自的信任度,再以这3类信任度为独立证据,采用D-S证据理论对3类证据进行融合,并依据分类判决规则得到最终的识别结果.试验结果表明,通过D-S理论对模糊向量机证据的融合,钨矿石初选的正确识别率达到96%以上,其准确率和稳定性较单一特征均有大幅度提高,满足生产过程中初选工艺的要求.According to the low accuracy and low stability of the single feature-based method for tungsten ore primary selection, a multi-feature fusion based on fuzzy support vector machine and D-S evidence theory was proposed. Firstly, the three types of vision feature that is color, gray and texture were extracted from the ore image after a series of image processing. Their probability function were acquired according to each type of feature utilizing fuzzy support vector machine and the results were used to D-S evidence theory as evidence. Finally, using D-S combination rule of evidence to achieve the decision fusion and giving final recognition result by classification rules. The experimental results show that the accuracy of multi-feature fusion methods is over 96% and it has good performance on accuracy and stability compared to the single feature-based method in tungsten ore primary selection. The accuracy and stability can meet the requirement of production process.

关 键 词：机器视觉 图像处理 D-S证据理论 钨矿石 模糊支持向量机 决策级融合 钨矿石初选 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]