基于香橙派Orange Pi 5B开发的矿石识别系统  

Ore Identification System Developed Based on Orange Pi 5B

作　　者：付琴 林奇兵 常周林[1] FU Qin;LIN Qibing;CHANG Zhoulin(School of Mechanical and Electrical Engineering,Guangdong Mechanical&.Electrical Polytechnic,Dongguan,Guangdong 523083,China)

机构地区：[1]广东科技学院机电工程学院,广东东莞市523083

出　　处：《矿业研究与开发》2025年第2期254-260,共7页Mining Research and Development

基　　金：广东省教育厅青年创新人才项目(2021KQNCX119);广东科技学院电子信息工程一流专业建设项目(GKZLGC2023156)。

摘　　要：针对目前矿石检测技术准确性不高、效率低、实时性差以及环境适应性弱等问题,提出了一种基于香橙派的矿石识别系统。该系统利用高清摄像头实时捕获矿石图像,并通过香橙派Orange Pi 5B开发板搭载YOLOv4-tiny-tf2模型处理图像数据,实现矿石种类的高效准确识别。结果表明:轻量化的YOLOv4-tiny-tf2模型在不牺牲模型精度的情况下简化了其结构,更适合在配置较低的设备上部署,具备了较高的便携性;相较于传统的人工视觉检测和化学分析方法,基于香橙派Orange Pi 5B的矿石识别系统能够实时识别矿石,识别准确率达到93.75%,具有较高的可靠性和稳定性。In response to the current challenges of low accuracy,inefficiency,poor real-time performance,and weak environmental adaptability in ore detection technology,a ore identification system based on Orange Pi was proposed.This system utilized a high-definition camera to capture real-time images of ores and processed image data using the YOLOv4-tiny-tf2 model on the Orange Pi 5B development board to achieve efficient and accurate identification of ore types.The results show that the lightweight YOLOv4-tiny-tf2 model simplifies its structure without sacrificing the accuracy of the model,which is more suitable for deployment on lower-configured devices and has a higher portability.Compared with the traditional artificial visual inspection and chemical analysis methods,the ore identification system based on Orange Pi 5B can identify the ore in real time,and the identification accuracy rate reaches 93.75%,which has high reliability and stability.

关 键 词：矿石识别系统 香橙派Orange Pi 5B YOLOv4-tiny-tf2 准确率 

分 类 号：TD803[矿业工程—矿山开采]