Heteroskedastic Geospatial Tracking with Distributed Camera Networks
This work addresses the problem of accurate and uncertain geospatial tracking for applications like surveillance or autonomous systems, but it is incremental as it builds on existing visual tracking methods with new data and constraints.
The paper tackled geospatial object tracking with distributed camera networks, predicting object tracks in geospatial coordinates with uncertainty while respecting communication constraints, and introduced a novel dataset and modeling framework that improved performance through uncertainty calibration and fine-tuning.
Visual object tracking has seen significant progress in recent years. However, the vast majority of this work focuses on tracking objects within the image plane of a single camera and ignores the uncertainty associated with predicted object locations. In this work, we focus on the geospatial object tracking problem using data from a distributed camera network. The goal is to predict an object's track in geospatial coordinates along with uncertainty over the object's location while respecting communication constraints that prohibit centralizing raw image data. We present a novel single-object geospatial tracking data set that includes high-accuracy ground truth object locations and video data from a network of four cameras. We present a modeling framework for addressing this task including a novel backbone model and explore how uncertainty calibration and fine-tuning through a differentiable tracker affect performance.