Diseño e implementación de algoritmos de visión artificial aplicados en vehículos aéreos no tripulados para aterrizaje autónomo. Caso de estudio: recolección de datos de sensores implementados en agricultura de precisión

Abstract

In networks that cover wide geographic areas, scalability and distance between nodes make it difficult to transfer data to information management points. Generally, these types of networks are used for monitoring and are deployed using a Wireless Sensor Network (WSN). Precision agriculture is a common and timely scenario for the deployment of these types of networks. These networks collect information from the environment, transmit it and process it. The low performance of the network nodes requires the use of various methods and structures that facilitate the transfer of information. A Delay Tolerant Network (DTN) is presented as a timely solution to ensure the transfer of information through store and forward. Within this scenario, a DTN can employ an unmanned aerial vehicle as a mule node to transport the information. The development and implementation of computer vision algorithms can improve the performance of the actions and processes developed by the unmanned vehicle. The present work addresses the mobile node landing stage as an alternative to improve data collection in WSNs using a DTN. The adjustment and customization provided by the autopilot allows total control of the movement and actions carried out by the vehicle. The difficulty in developing algorithms for this type of system lies in the possible errors that the vehicle experiences before and during the flight. Although the calibration and configuration of the vehicle provide a degree of confidence in its operation, ArduPilot does not guarantee that errors will not occur due to either the software or the hardware. Therefore, it is essential to take into account the challenges linked to the behavior of the vehicle and its environment. This degree work is a first approach for the development of applications that exploit the adaptability of an unmanned aerial vehicle in applications that deploy WSNs. In this first approach, it was possible to use a camera to obtain characteristics of a marker ArUco, and use them to design algorithms that control the movement of the drone in the landing stage and data collection, and evaluate the impact of the environment on the algorithms that make up the landing system. In addition, it was possible to introduce landing as an alternative to reduce energy consumption in data collection using drones, demonstrate the operation of a DTN, consider it as a valid option for data collection in WSNs, and determine the main limitations present in the development of applications involving unmanned aerial vehicles.