摘要
矿山异物检测是异物智能化去除的前提,更是保障设备安全运行,矿山正常生产的关键。在矿山生产过程中,异物来源广泛,种类繁杂。针对传统的矿山异物检测方法面临适应性差和效率低的问题,提出了一种面向大型振动筛筛面的矿山异物检测算法模型。为解决强振动、矿石遮挡和粉尘水雾等复杂环境的干扰,该模型引入了改进的显式视觉中心模块(EVCBlock),轻量化上采样算子CARAFE和基于动态非单调聚焦机制的梯度增益损失函数WiseIoU-v3,有效提升了在复杂环境下的异物检测性能。利用TensorRT对模型优化并部署至边缘计算设备Jetson Xavier NX,实现了在边缘侧的异物检测。研究结果表明:该模型在振动筛筛面异物检测上的表现明显好于其他对比模型。经多线程视频推流测试,模型部署至边缘计算设备平均识别精确率可以达到96.3%,平均帧率达到25 FPS以上,满足了实际检测要求。
Foreign bodies detection in mines is the premise of intelligent removal of foreign bodies,but also the key to ensure the safe operation of equipment and the normal production of the diggings.Foreign bodies come from a wide range of sources and various types in the process of mines production.Aiming at the problems of poor adaptability and low efficiency of traditional foreign bodies detection methods,a foreign bodies detection algorithm model for surface of large vibrating screen was proposed.In order to solve the interference of complex environment such as violent vibration,shielding of ore,dust and water mist,the improved Explicit Vision Center block(EVCBlock),lightweight up-sampling operator CARAFE and gradient gain loss function WiseIoU-v3 based on dynamic non-monotony focusing mechanism are introduced in this model,which effectively improve the detection performance of foreign bodies in complex environment.The model with TensorRT optimization was deployed to the edge computing device Jetson Xavier NX to achieve foreign object detection on the edge side.The results show that the proposed model is better than other models in detecting foreign bodies on the vibrating screen surface.After multithreaded video push streaming test,the average accuracy of deploying to the edge computing device can reach 96.3%,and the average frame rate can reach more than 25 FPS,which meets the actual detection requirements.
作者
刘善明
余新阳
欧阳魁
LIU Shan-ming;YU Xin-yang;OUYANG Kui(Jiangxi Province Key Laboratory of Mining Engineering,Ganzhou 341000,China;School of Resources and Environmental Engineering,Jiangxi University of Science and Technology,Ganzhou 341000,China;Hunan PDC Mining Technology Co.,Ltd.,Changsha 410000,China)
出处
《计算机技术与发展》
2024年第5期196-204,共9页
Computer Technology and Development
基金
国家自然科学基金(52264023)。
