Early development of horizon detection system based on OpenCV and Python

To stabilize and control a flying vehicle, attitude sensors are needed. Relative attitude sensors such as gyroscope tend to accumulate errors if not corrected periodically using absolute attitude sensors. Horizon sensor is one of the absolute sensors that is commonly used in satellites and recently...

Full description

Saved in:
Bibliographic Details
Main Author: Adi, Andreas P.
Format: Conference Proceeding
Language:English
Subjects:
Algorithms
Attitudes
Edge detection
Horizon
Parameters
Pitch (inclination)
Rolling motion
Sensors
Straight lines
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To stabilize and control a flying vehicle, attitude sensors are needed. Relative attitude sensors such as gyroscope tend to accumulate errors if not corrected periodically using absolute attitude sensors. Horizon sensor is one of the absolute sensors that is commonly used in satellites and recently in autonomous UAVs. This paper describes our early development of a horizon detection system, assuming the horizon as a straight line. The input image is pre-processed by converting to grayscale and applying a Gaussian filter. Then the Canny method performs edge detection and the detected edges are given to the line fitting algorithm to find the best fit line equation. We use the least square method to do line fitting. From the line equation obtained, roll and pitch angles are calculated. We implement the system using OpenCV and Python. We have conducted lab-scale experiments using images and video files. We make parameter adjustments to obtain good results. For the images we used, the recommended parameter values are Canny threshold (200, 250) and kernel size 3x3, and Gaussian filter size 3x3 to 5x5 in the Pre-processing stage. For other images, the values may need to be adjusted accordingly.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0060250