Image Processing Based Drone Landing Technique Considering GPS error and wind Direction

Abstract

The fourth industrial revolution represents the transition to an intelligent society that combines sensor, network, and artificial intelligence to significantly improve human life. Recently emerged drones are replacing manned airplanes, which are expensive to maintain, and are used in a variety of fields such as shooting, shipping, and surveillance. Methods/Statistical analysis: Drones are required to perform the duties based on accurate location. Therefore, it is required to identify an exact location. However, drones are using a position sensor, GPS, but still suffering from errors including the location of satellite, transmission error, and internally installed clock in the satellite causing an error with several meters. In this study, image process-based drone landing techniques are researched for correcting the location of drone in the use of drone camera that is used to supplement GPS. Findings: In this study, camera was connected to Raspberry Pi that was used as a drone controller to record while analyzing photos recorded in a certain period of time to supplement the error in GPS estimates to find target point. Suggested drone landing technique is programmed to find an exact landing area by determining the moving direction of drones towards the wind measured with position sensor in flight controller and the direction of wind estimated with image if there was influence of wind. Due to various changes in environment in daily lives, experiment was conducted in Gazebo simulator of a drone. Improvements/Applications: It was confirmed that drones were able to access to the error within 1.5m from pre-determined landing area in spite of the influence of wind.