Presentation Title

Autonomous Navigation and Localization of Unmanned Vehicles using Computer Vision

Presenter Information

Travis GraftonFollow

Faculty Mentor

Dr. Alec Sim

Start Date

17-11-2018 8:30 AM

End Date

17-11-2018 10:30 AM

Location

HARBESON 63

Session

POSTER 1

Type of Presentation

Poster

Subject Area

engineering_computer_science

Abstract

Unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) are expected to be used in future missions to Mars. The OpenCV Python library enables the use of computer vision and machine learning applications. These applications are a viable candidate for autonomous localization and navigation of UAVs and UGVs. The goal of this study is to use the OpenCV library to teach UAVs and UGVs to identify ground features likely to be seen on a future Mars Mission. Haar Cascades are used to identify circles as potential craters. Data from the Robbins Crater Database, which lists information relating to approximately 640,000 craters on the surface of Mars, is used in conjunction with the Keras Deep Learning Library to train a network to identify images as craters. A database was developed to train a neural network using the YOLO (You Only Look Once) algorithm, which feeds bounded images into a neural net. The trained model will be able to both recognize craters and give a relative position in the image by bounding the crater with a box. Real time processing will require the integration of a mobile camera system, PixHawk autopilot system, and ODroid microcomputer.

This document is currently not available here.

Share

COinS
 
Nov 17th, 8:30 AM Nov 17th, 10:30 AM

Autonomous Navigation and Localization of Unmanned Vehicles using Computer Vision

HARBESON 63

Unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) are expected to be used in future missions to Mars. The OpenCV Python library enables the use of computer vision and machine learning applications. These applications are a viable candidate for autonomous localization and navigation of UAVs and UGVs. The goal of this study is to use the OpenCV library to teach UAVs and UGVs to identify ground features likely to be seen on a future Mars Mission. Haar Cascades are used to identify circles as potential craters. Data from the Robbins Crater Database, which lists information relating to approximately 640,000 craters on the surface of Mars, is used in conjunction with the Keras Deep Learning Library to train a network to identify images as craters. A database was developed to train a neural network using the YOLO (You Only Look Once) algorithm, which feeds bounded images into a neural net. The trained model will be able to both recognize craters and give a relative position in the image by bounding the crater with a box. Real time processing will require the integration of a mobile camera system, PixHawk autopilot system, and ODroid microcomputer.