基于YOLOV4的港口作业人员检测系统研究

随着经济高速发展,经略海洋与海洋经济发展成为国家重要发展战略,港口设施建设和完善对于海洋经济发展有着重要作用。然而,港口重型机械以及轮机操作对作业人员安全有潜在威胁,智能化、信息化码头建设势在必行。近年计算机视觉和深度学习技术快速发展,为港口应用智能视觉技术提供了有力的技术支撑。基于深度学习框架YOLOV4搭建了港口作业人员目标检测平台,在自建港口收集并整理了一个大规模作业人员视频数据集,在该数据集上实现不同作业场景下港口作业人员的精确检测。在自建港口作业人员数据集上将FasterRCNN、SSD和YOLOV4三种目标检测框架进行实验对比,结果表明,YOLOV4的平均检测准确率优于其它目标检测框架。基于YOLOV4的港口作业人员检测系统应用提高了港口信息化建设进度,提高了港口作业人员的安全性。

Photoluminescence from neodymium silicide thin films formed by MEVVA ion source

Neodymium silicides were synthesized by Nd ion implant6d into Si substrates with the aid of a metal vapor vacuum arc (MEVVA) ion source. The blender of Nd5Si4 and NdSi2 was formed in a neodymium-implanted silicon thin film during the as-implanted state, but there was only single neodymium silicide compound in the postannealed state, and the phase changed from NdSi2 to Nd5Si4 with increasing annealing temperature. The blue-violet luminescence excited by ultra-violet was observed at the room temperature (RT), and the intensity of photoluminescence (PL) increased with increasing the neodymium ion fluence. Moreover, the photoluminescence was closely dependent on the temperature of rapid thermal annealing (RTA). A mechanism of photoluminescence was discussed.