giant.ufo.ekf_tracker

This module provides a class for tracking UFO detections across monocular images.

Description

UFO tracking is done using Extended Kalman Filters to predict the locations of UFOs in subsequent images and then match UFO detections in those images to the predict locations. Multiple paths are followed for each possible particle resulting in many possible tracks across all of the images. Each of these possible tracks is then filtered to remove extraneous tracks (and tracks that are subsets of other tracks) leaving only the tracks we are relatively confident in the quality of. Further description of the process can be found in the paper at https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2019EA000843

Use

To track UFOs from image to image, initialize the Tracker class with the required information and then call method Tracker.track(). This will go through the whole process of initial tracking and then filtering for all the possible particles in the loaded data, storing the results in the confirmed_tracks and confirmed_standard_deviations lists. You can also save the results to a csv file using method save_results().

You may also be interested in using the UFO class which combines both detection and tracking instead of using this class directly.

Classes

Tracker

This class provides an interface for autonomously tracking UFOs through subsequent images in time.