铜业工程Issue(2)：48-54,7.DOI:10.3969/j.issn.1009-3842.2025.02.006
基于EffCD-YOLOv8的矿物检测识别技术研究
Mineral Detection and Identification Technology Based on EffCD-YOLOv8
摘要
Abstract
Due to the diversity of mineral types and the large quantity of minerals mined at one time,using visual recognition technolo-gy for classification often leads to missed and incorrect detections.To achieve accurate recognition,annotation,and classification of minerals,this paper proposes EffCD-YOLOv8 algorithm based on YOLOv8 network structure.The algorithm integrates EfficientNetv2 module into the backbone network,enhancing detection accuracy while maintaining training speed.CCFF module is incorporated into the neck to achieve cross-scale feature fusion,and DBB module is utilized in the head to process objects across different spatial scales,enabling more precise recognition,annotation,and classification.The results demonstrate that EffCD-YOLOv8 model achieves a recog-nition accuracy of 83.2%on a custom mineral dataset,outperforming the original YOLOv8 algorithm and effectively enabling accurate recognition,annotation,and classification of minerals.关键词
矿石分类/YOLOv8/视觉识别技术/目标检测Key words
ore classification/YOLOv8/visual recognition technology/target detection分类
冶金工业引用本文复制引用
曾赫,肖罡,蔺永诚..基于EffCD-YOLOv8的矿物检测识别技术研究[J].铜业工程,2025,(2):48-54,7.基金项目
国家重点研发计划项目(2023YFC3904200) （2023YFC3904200）
江西省自然科学基金项目(20224ACB218002) （20224ACB218002）
江西省重大科技研发专项项目(20232ACE01010)资助 （20232ACE01010）
