期刊文献+

基于RetinaNet深度网络的煤矿带式输送机异物智能识别方法

Intelligent identification method of foreign objects on coal mine belt conveyor based on RetinaNet deep network
下载PDF
导出
摘要 为解决煤矿带式输送机异物识别的尺度变化问题,提高煤矿生产安全性和效率,提出基于RetinaNet深度网络的煤矿带式输送机异物智能识别方法。通过局部伽马变换和单参数同态滤波对摄像机采集的煤矿带式输送机图像进行增强处理,为后续异物识别提供良好基础。将增强后的图像输入到RetinaNet深度网络中,通过其内部特征金字塔网络提取煤矿带式输送机运输图像深度特征,输入到各子网中实现边框回归和分类,同时定义损失函数对网络进行调整。通过大量包含煤矿带式输送机异物的图像来训练RetinaNet深度网络,最终通过训练好的网络实现煤矿带式输送机异物智能识别。通过实验验证,该方法能够迅速且准确地识别出异物,并通过醒目红色标识框对异物进行精确标注,提升带式输送机运输的连续性和稳定性,有效提高煤矿的生产效率和经济效益。 In order to solve the scale change problem of foreign objects identification and improve the safety and efficiency of coal mine production,an intelligent identification method for foreign objects on coal mine belt conveyor based on the RetinaNet deep network was proposed.Local gamma transform and single parameter homomorphic filter were used to enhance the coal mine belt conveyor image,which provided a good basis for the subsequent foreign objects identification.Inputting the enhanced images into the RetinaNet deep network,the depth features of coal mine belt conveyor images were extracted through its internal feature pyramid network,and border regression and classification were achieved through inputting them into each subnet,and loss functions were defined to adjust the network.The RetinaNet deep network was trained through a large number of images containing foreign objects on the belt conveyor,and intelligent identification of foreign objects on the belt conveyor was finally realized through the trained network.The experimental verified that the method could quickly and accurately identify foreign bodies,and accurately mark foreign bodies through an eye-catching red label box,improve the continuity and stability of belt conveyor transportation,and effectively improve the production efficiency and economic benefits of coal mines.
作者 丁文博 云龙 DING Wenbo;YUN Long(Yujialiang Coal Mine,China Energy Shendong Coal Group Co.,Ltd.,Yulin,Shaanxi 719315,China)
出处 《中国煤炭》 北大核心 2024年第S01期75-81,共7页 China Coal
关键词 RetinaNet深度网络 煤矿带式输送机运输 异物智能识别 局部伽马变换 单参数同态滤波 特征金字塔网络 RetinaNet deep network coal mine belt conveyor transportation intelligent identification of foreign bodies local gamma transform single parameter homomorphic filter feature pyramid network
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部