Skip to content

Autonomous object tracking: A combination of a detector and a tracker

License

Notifications You must be signed in to change notification settings

COVISART/detection-and-tracking

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Detection and Tracking

Repository for Master's Thesis in Engineering Cybernetics at NTNU, 2018.

Title: Autonomous Target Detection and Tracking for Remotely operated Weapon Stations

Intention

Detect and track targets of interest in camera video, for a Remote Weapon Station (RWS).

Approach

Combine an accurate detector with a fast tracker. Methods of interest:

  • Detector based on deep learning
  • Point based tracker

Motivation/Design

If a detector process frames slower than the video frame rate, a number of frames will be skipped in between detections - and the detection will also be a few frames "old" when presented. A solution to this is to buffer the skipped frames, and retrace this buffer with a fast enough tracker, to make the detection relevant for the current frame.

The implementation host/tracking-app.pyis an autonomous tracker - a real-time tracker with periodic corrections from a deep learning detector.

Software Framework

Hardware

  • Host: Desktop PC (optional CPU/GPU depending on TensorFlow distribution)
  • Target: Nvidia Jetson TX (future work)

About

Autonomous object tracking: A combination of a detector and a tracker

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 99.0%
  • Other 1.0%