Recent advances in the research of autonomous vehicles have showed a vast range of applications, such as exploration, surveillance and environmental monitoring. Considering the mining industry, it is possible to use such vehicles in the prospection of minerals of commercial interest beneath the ground. However, tasks such as geophysical surveys are highly dependent on specific sensors, which mostly are not designed to be used in these new range of autonomous vehicles. In this work, we propose a novel magnetic survey pipeline that aims to increase versatility, speed and robustness by using autonomous rotary-wing Unmanned Aerial Vehicles (UAVs). We also discuss the development of a state-of-the-art three-axis fluxgate, where our goal in this work was to refine and adjust the sensor topology and coupled electronics specifically for this type of vehicle and application. The sensor was built with two ring-cores using a specially developed stress-annealed CoFeSiB amorphous ribbon, in order to get sufficient resolution to detect concentrations of small ferrous minerals. Finally, we report on the results of experiments performed with a real UAV in an outdoor environment, showing the efficacy of the methodology in detecting an artificial ferrous anomaly.

We introduce a lightweight and accurate architecture for resource-efficient visual correspondence. Our method, dubbed XFeat (Accelerated Features), revisits fundamental design choices in convolutional neural networks for detecting, extracting, and matching local features. Our new model satisfies a critical need for fast and robust algorithms suitable to resource-limited devices. In particular, accurate image matching requires sufficiently large image resolutions - for this reason, we keep the resolution as large as possible while limiting the number of channels in the network. Besides, our model is designed to offer the choice of matching at the sparse or semi-dense levels, each of which may be more suitable for different downstream applications, such as visual navigation and augmented reality. Our model is the first to offer semi-dense matching efficiently, leveraging a novel match refinement module that relies on coarse local descriptors. XFeat is versatile and hardware-independent, surpassing current deep learning-based local features in speed (up to 5x faster) with comparable or better accuracy, proven in pose estimation and visual localization. We showcase it running in real-time on an inexpensive laptop CPU without specialized hardware optimizations. Code and weights are available at www.verlab.dcc.ufmg.br/descriptors/xfeat_cvpr24.