Tracking moving objects is a relevant topic in a wide range of research activities. To meet this goal, sensor fusion is a frequently used approach that is able to exploit the synergies from different sensors, achieving high precision and accuracy. In this article, an application of this technique is presented for an actual industrial scenario, namely, the shield metal arc welding (SMAW) process. A wearable real-time tracking system, employing a camera and an inertial unit on a commercial helmet, is described and assessed to estimate the speed of a light-emitting target, namely, the welding pool, despite the noises introduced by sparks, smoke, and high-density illumination. The vision-based module was used to follow the weld pool and to calibrate automatically the system, while the inertial one was used to compensate for the influence of the yaw-angular displacements. Results from tests performed on four batches of electrodes showed how the system is able to catch the phenomenon with errors less than 25%. Compared with other systems, our prototype is able to acquire data from moving objects in real time through sensors positioned onto a moving reference system.

中文翻译：

---

用于融合视觉和惯性传感器的移动发光目标的头戴式独立实时跟踪系统


跟踪移动物体是广泛研究活动中的一个相关主题。为了实现这一目标，传感器融合是一种常用的方法，它能够利用不同传感器的协同作用，实现高精度和准确度。在本文中，介绍了该技术在实际工业场景中的应用，即屏蔽金属电弧焊（SMAW）工艺。描述和评估了一种可穿戴实时跟踪系统，该系统在商用头盔上采用摄像头和惯性单元，以估计发光目标（即焊接熔池）的速度，尽管存在火花、烟雾、和高密度照明。基于视觉的模块用于跟踪熔池并自动校准系统，而惯性模块用于补偿偏航角位移的影响。对四批电极进行的测试结果表明，该系统能够以低于 25% 的误差捕获这种现象。与其他系统相比，我们的原型能够通过位于移动参考系统上的传感器实时获取移动物体的数据。